├── README.md ├── about.md ├── blog ├── chat-api.md ├── deep-research.md ├── introducing-basis-with-calibrated-confidences.md ├── parallel-search-api.md ├── parallel-search-mcp-in-devin.md ├── parallel-task-api.md ├── search-api-benchmark.md ├── search-mcp-server.md ├── source-policy.md └── task-group-api.md ├── fetch.js ├── platform.parallel.ai └── pricing.md ├── privacy-policy.md ├── terms-of-service.md └── urls.md /README.md: -------------------------------------------------------------------------------- 1 | Scraped from [urls.md](urls.md) using r.jina.ai 2 | 3 | Context: [![](https://badge.forgithub.com/janwilmake/parallel-website)](https://uithub.com/janwilmake/parallel-website) 4 | 5 | Missing: 6 | 7 | - Careers: https://jobs.ashbyhq.com/parallel (captcha hit) 8 | -------------------------------------------------------------------------------- /about.md: -------------------------------------------------------------------------------- 1 | # About 2 | 3 | The world wide web is built for humans using browsers and apps. Soon, the web will be used primarily by AIs on our behalf. While people typically search and discover common information with a limited number of access patterns, AIs can unlock immense value from the long-tail of specialized knowledge on the web. As a result, access and computation on the web will increase by several orders of magnitude. Everything from the infrastructure to business models will need to evolve. 4 | 5 | At Parallel Web Systems, we are bringing this new web to life: it’s built with, by, and for AIs. Our work spans innovations across crawling, indexing, ranking, retrieval, and reasoning systems. Our first product is an API for AIs to execute high-value repetitive tasks using web data. 6 | Reach out at hello@parallel.ai. 7 | -------------------------------------------------------------------------------- /blog/chat-api.md: -------------------------------------------------------------------------------- 1 | Title: Introducing the Parallel Chat API 2 | 3 | URL Source: https://parallel.ai/blog/chat-api 4 | 5 | Published Time: 2025-05-30T00:23:35Z 6 | 7 | Markdown Content: 8 | Introducing the Parallel Chat API | Enterprise Deep Research API for ChatGPT & Claude 9 | 10 | =============== 11 | 12 | [Parallel](https://parallel.ai/)[About](https://parallel.ai/about)[About](https://parallel.ai/about)[Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog)[Docs](https://docs.parallel.ai/introduction/quickstart)[Docs](https://docs.parallel.ai/introduction/quickstart) 13 | 14 | [Start Building P](https://platform.parallel.ai/)[Start Building](https://platform.parallel.ai/) 15 | 16 | Menu[Menu] 17 | 18 | Human Machine 19 | 20 | #Introducing the Parallel Chat API 21 | ================================== 22 | 23 | Starting today, the Parallel Chat API is generally available in beta - bringing real-time web research to interactive AI applications using our rapidly growing web index. 24 | 25 | Tags:Product Release 26 | 27 | Reading time:1 min 28 | 29 | ![Image 1: Introducing the Parallel Chat API - a low latency web research API for web based LLM completions. The Parallel Chat API returns completions in text and structured JSON format, and is OpenAI Chat Completions compatible. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F780ef9d2c09f2e3526a0d11a4bda9eb0a45ef391-2560x1704.jpg&w=3840&q=75) 30 | 31 | The Chat API returns OpenAI ChatCompletions compatible streaming text and JSON with the low latency your interactive apps need. It is built with the same obsession for price-performance that we’ve demonstrated with the [Task API](https://parallel.ai/blog/parallel-task-api)[Task API]($https://parallel.ai/blog/parallel-task-api). 32 | 33 | ###Python Code Snippet 34 | 35 | 1 36 | 37 | 2 38 | 39 | 3 40 | 41 | 4 42 | 43 | 5 44 | 45 | 6 46 | 47 | 7 48 | 49 | 8 50 | 51 | 9 52 | 53 | 10 54 | 55 | 11 56 | 57 | 12 58 | 59 | 13 60 | 61 | 14 62 | 63 | 15 64 | 65 | 16 66 | 67 | 17 68 | 69 | 18 70 | 71 | 19 72 | 73 | 20 74 | 75 | 21 76 | 77 | 22 78 | 79 | 23 80 | 81 | 24 82 | 83 | 25 84 | 85 | 26 86 | 87 | 27 88 | 89 | 28 90 | 91 | 29 92 | 93 | 30 94 | 95 | 31 96 | 97 | 32 98 | 99 | 33 100 | 101 | 34 102 | 103 | 35 104 | 105 | 36 106 | 107 | 37 108 | 109 | 38 110 | 111 | 39 112 | 113 | ``` 114 | from openai import OpenAI 115 | 116 | client = OpenAI( 117 | api_key="PARALLEL_API_KEY", # Your Parallel API key 118 | base_url="https://beta.parallel.ai" # Parallel's API beta endpoint 119 | ) 120 | 121 | response = client.chat.completions.create( 122 | model="speed", # Parallel model name 123 | messages=[ 124 | {"role": "user", "content": "What does Parallel Web Systems do?"} 125 | ], 126 | response_format={ 127 | "type": "json_schema", 128 | "json_schema": { 129 | "name": "reasoning_schema", 130 | "schema": { 131 | "type": "object", 132 | "properties": { 133 | "reasoning": { 134 | "type": "string", 135 | "description": "Think step by step to arrive at the answer", 136 | }, 137 | "answer": { 138 | "type": "string", 139 | "description": "The direct answer to the question", 140 | }, 141 | "citations": { 142 | "type": "array", 143 | "items": {"type": "string"}, 144 | "description": "Sources cited to support the answer", 145 | }, 146 | }, 147 | }, 148 | }, 149 | }, 150 | ) 151 | 152 | print(response.choices[0].message.content) 153 | ``` 154 | ```` from openai import OpenAI client = OpenAI( api_key="PARALLEL_API_KEY", # Your Parallel API key base_url="https://beta.parallel.ai" # Parallel's API beta endpoint) response = client.chat.completions.create( model="speed", # Parallel model name messages=[ {"role": "user", "content": "What does Parallel Web Systems do?"} ], response_format={ "type": "json_schema", "json_schema": { "name": "reasoning_schema", "schema": { "type": "object", "properties": { "reasoning": { "type": "string", "description": "Think step by step to arrive at the answer", }, "answer": { "type": "string", "description": "The direct answer to the question", }, "citations": { "type": "array", "items": {"type": "string"}, "description": "Sources cited to support the answer", }, }, }, }, },) print(response.choices[0].message.content)```` 155 | 156 | Last year, we started building our own web crawler and index, recognizing that AIs need native infrastructure to query and reason over the web. The Chat API answers instantly using our index, while the Task API combines our index with real-time web crawling for maximum accuracy and the freshest data. Both APIs return comprehensive citations so you can always verify responses. 157 | 158 | ![Image 2: Illustration demonstrating deep research API concepts, web search capabilities, or AI agent integration features](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fdfe299127abfeba88c55260ef84f9a465d733a65-1572x1080.gif&w=3840&q=75) 159 | 160 | ![](https://cdn.sanity.io/images/5hzduz3y/production/dfe299127abfeba88c55260ef84f9a465d733a65-1572x1080.gif) 161 | 162 | ## 163 | 164 | ****Start building.**** 165 | --------------------------- 166 | 167 | Get started in our [Developer Platform](https://platform.parallel.ai/play/chat)[Developer Platform]($https://platform.parallel.ai/play/chat) or dive into the [documentation](https://docs.parallel.ai/resources/chat-api)[documentation]($https://docs.parallel.ai/resources/chat-api). 168 | 169 | ![Image 3: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) 170 | 171 | By Parallel 172 | 173 | May 30, 2025 174 | 175 | ### ##Related Posts 10 176 | 177 | [![Image 4: A linear dithering of a search interface for agents](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fcf3d3b4112c8ab27af1a56442e549a6c9fb2b220-1600x1064.png&w=3840&q=75) ![Image 5: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[A state-of-the-art search API purpose-built for agents](https://parallel.ai/blog/search-api-benchmark) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/search-api-benchmark) 178 | 179 | [![Image 6: Parallel Search MCP Server in Devin](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fdfc00df5061085ec771f28d263b231d4c53747c3-1600x1044.png&w=3840&q=75) ![Image 7: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Parallel Search MCP Server in Devin](https://parallel.ai/blog/parallel-search-mcp-in-devin) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/parallel-search-mcp-in-devin) 180 | 181 | [![Image 8: Introducing Tool Calling via MCP Servers](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fe220ba8a89dfaeb943f943fc92636629a0b9ba14-1600x1600.png&w=3840&q=75) ![Image 9: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Tool Calling via MCP Servers](https://parallel.ai/blog/mcp-tool-calling) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/mcp-tool-calling) 182 | 183 | [![Image 10: Introducing the Parallel Search MCP Server ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F32cf3f1026d452dbaed58e8df4d7b487166d423a-1972x1972.png&w=3840&q=75) ![Image 11: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Search MCP Server ](https://parallel.ai/blog/search-mcp-server) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/search-mcp-server) 184 | 185 | [![Image 12: Starting today, Source Policy is available for both the Parallel Task API and Search API - giving you granular control over which sources your AI agents access and how results are prioritized.](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9eec7c7314218135df8fd7a83b1c6d3b953eece9-1528x1016.png&w=3840&q=75) ![Image 13: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Source Policy](https://parallel.ai/blog/source-policy) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/source-policy) 186 | 187 | [![Image 14: The Parallel Task Group API](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F5b63138a650248d4454f306ba72d99dc1572e08a-1600x946.png&w=3840&q=75) ![Image 15: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[The Parallel Task Group API](https://parallel.ai/blog/task-group-api) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/task-group-api) 188 | 189 | [![Image 16: State of the Art Deep Research APIs](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F897a6d4a4d5caf5657b59cdc1dc99de0641c66df-1600x900.png&w=3840&q=75) ![Image 17: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[State of the Art Deep Research APIs](https://parallel.ai/blog/deep-research) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/deep-research) 190 | 191 | [![Image 18: Introducing the Parallel Search API ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Ff61c8461515390c7e546ee3ac1b470a4f79084cc-2560x1704.png&w=3840&q=75) ![Image 19: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Search API ](https://parallel.ai/blog/parallel-search-api) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/parallel-search-api) 192 | 193 | [![Image 20: Parallel Web Systems introduces Basis with calibrated confidences - a new verification framework for AI web research and search API outputs that sets a new industry standard for transparent and reliable deep research. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F20273a28dbb0a5ee55784f4c9f7050427094f4eb-2672x1512.png&w=3840&q=75) ![Image 21: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Basis with Calibrated Confidences ](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) Tags:Product Release Reading time:4 min](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) 194 | 195 | [![Image 22: The Parallel Task API is a state-of-the-art system for automated web research that delivers the highest accuracy at every price point. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F96491c652ee4988fd9866d1b7619407582879e64-2576x1448.png&w=3840&q=75) ![Image 23: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Task API](https://parallel.ai/blog/parallel-task-api) Tags:Product Release,Benchmarks Reading time:4 min](https://parallel.ai/blog/parallel-task-api) 196 | 197 | ![Company Logo](https://parallel.ai/parallel-logo-540.png) 198 | 199 | ### Company 200 | 201 | * [hello@parallel.ai](mailto:hello@parallel.ai)[hello@parallel.ai](mailto:hello@parallel.ai) 202 | 203 | ### Resources 204 | 205 | * [About](https://parallel.ai/about)[About](https://parallel.ai/about) 206 | * [Docs](https://docs.parallel.ai/)[Docs](https://docs.parallel.ai) 207 | * [Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog) 208 | * [Changelog](https://docs.parallel.ai/resources/changelog)[Changelog](https://docs.parallel.ai/resources/changelog) 209 | * [Careers](https://jobs.ashbyhq.com/parallel)[Careers](https://jobs.ashbyhq.com/parallel) 210 | 211 | ### Info 212 | 213 | * [Terms](https://parallel.ai/terms-of-service)[Terms](https://parallel.ai/terms-of-service) 214 | * [Privacy](https://parallel.ai/privacy-policy)[Privacy](https://parallel.ai/privacy-policy) 215 | * [Trust Center](https://trust.parallel.ai/)[Trust Center](https://trust.parallel.ai/) 216 | 217 | ![SOC 2 Compliant](https://parallel.ai/soc2.svg) 218 | 219 | [LinkedIn](https://www.linkedin.com/company/parallel-web/about/)[LinkedIn](https://www.linkedin.com/company/parallel-web/about/) 220 | 221 | Parallel Web Systems Inc. 2025 222 | -------------------------------------------------------------------------------- /blog/deep-research.md: -------------------------------------------------------------------------------- 1 | Title: State of the Art Deep Research APIs 2 | 3 | URL Source: https://parallel.ai/blog/deep-research 4 | 5 | Published Time: 2025-06-17T16:24:26Z 6 | 7 | Markdown Content: 8 | State of the Art Deep Research APIs | Enterprise Deep Research API for ChatGPT & Claude 9 | 10 | =============== 11 | 12 | [Parallel](https://parallel.ai/)[About](https://parallel.ai/about)[About](https://parallel.ai/about)[Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog)[Docs](https://docs.parallel.ai/introduction/quickstart)[Docs](https://docs.parallel.ai/introduction/quickstart) 13 | 14 | [Start Building P](https://platform.parallel.ai/)[Start Building](https://platform.parallel.ai/) 15 | 16 | Menu[Menu] 17 | 18 | Human Machine 19 | 20 | #State of the Art Deep Research APIs 21 | ==================================== 22 | 23 | Tags:Benchmarks 24 | 25 | Reading time:3 min 26 | 27 | ![Image 2: State of the Art Deep Research APIs](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F897a6d4a4d5caf5657b59cdc1dc99de0641c66df-1600x900.png&w=3840&q=75) 28 | 29 | Parallel Task API processors achieve state-of-the-art performance on [BrowseComp](https://openai.com/index/browsecomp/)[BrowseComp]($https://openai.com/index/browsecomp/), a challenging benchmark built by OpenAI to test web search agents' deep research capabilities. Our best processor reaches 27% accuracy— higher than the accuracy achieved by humans given 2 hours per problem. 30 | 31 | ##The deep research challenge 32 | ----------------------------- 33 | 34 | BrowseComp represents a new class of research problems that resist conventional web search. Unlike simple fact retrieval, these 1,266 questions require multi-hop reasoning across scattered sources, creative search reformulation when initial strategies fail, and synthesis of contextual clues spanning multiple time periods. 35 | 36 | Consider this sample question: 37 | 38 | ### Notes 39 | 40 | _\_"A piece of art was funded by a certain organization, according to an entry made on January 28, 2019. This piece of art belongs to an art form that has the support and acceptance of the local community, according to the organization's founder, as stated in a blog post from 2016. The artist who created the piece works under an alias, faced tough challenges growing up, features circles in their work often, and is fascinated by human behavior, according to another entry posted by the same organization from 2012. What's the title of the entry from 2019, as it appears on the organization's website?"\__ 41 | 42 | Human experts solve only about 25% of these questions correctly within two hours. While esoteric, they mirror critical business challenges that demand sophisticated needle-in-haystack capabilities: connecting regulatory filings across time periods for due diligence, synthesizing competitive intelligence from fragmented sources, tracking supply chain dependencies through multiple corporate layers, or conducting comprehensive background research where a single overlooked detail can derail major decisions. 43 | 44 | These are the research tasks that matter most to organizations—complex, multi-faceted investigations that traditional search tools handle poorly but that can make or break strategic initiatives. 45 | 46 | ##State of the art results 47 | -------------------------- 48 | 49 | Parallel Task API processors outperform human experts and all commercially available web search and deep research APIs on BrowseComp, while being significantly cheaper. 50 | 51 | BrowseComp 52 | 53 | [COST (CPM) ACCURACY (%) Loading chart...](https://parallel.ai/blog/deep-research) 54 | 55 | CPM: USD per 1000 requests. Cost is shown on a Linear scale. 56 | 57 | Parallel 58 | 59 | Others 60 | 61 | BrowseComp benchmark analysis: CPM: USD per 1000 requests. Cost is shown on a Linear scale. . Evaluation shows Parallel's enterprise deep research API for AI agents achieving up to 48% accuracy, outperforming GPT-4 browsing (1%), Claude search (6%), Exa (14%), and Perplexity (8%). Enterprise-grade structured deep research performance across Cost (CPM) and Accuracy (%). State-of-the-art enterprise deep research API with structured data extraction built for ChatGPT deep research and complex multi-hop AI agent workflows. 62 | 63 | ### ###About the benchmark 64 | 65 | This benchmark, created by OpenAI, contains 1,266 questions requiring multi-hop reasoning, creative search formulation, and synthesis of contextual clues across time periods. Read our blog [here](https://parallel.ai/blog/deep-research)[here]($https://parallel.ai/blog/deep-research). 66 | 67 | ### ###Steps of reasoning 68 | 69 | 100% Multi-Hop questions 70 | 71 | ### ###Parallel 600/1200 72 | 73 | Parallel 600 and 1200 are agents with the same architecture as Parallel Ultra (which costs 300 USD for 1000 queries), but with 2x and 4x the compute and cost. 74 | 75 | ###BrowseComp Scaled Compute Benchmark 76 | -------------------------------------- 77 | 78 | | Model | Cost (CPM) | Accuracy (%) | 79 | | ------------------------- | ---------- | ------------- | 80 | | Base | 10 | 4 | 81 | | Core | 25 | 7 | 82 | | Pro | 100 | 17 | 83 | | Ultra | 300 | 27 | 84 | | Parallel 600 | 600 | 39 | 85 | | Parallel 1200 | 1200 | 48 | 86 | | GPT-4.1 w/ browsing | 53 | 1 | 87 | | Claude Sonnet 4 w/ search | 1168 | 6 | 88 | | Exa Research | 275 | 14 | 89 | | Perplexity Deep Research | 880 | 8 | 90 | 91 | CPM: USD per 1000 requests. Cost is shown on a Linear scale. 92 | 93 | ### ###About the benchmark 94 | 95 | This benchmark, created by OpenAI, contains 1,266 questions requiring multi-hop reasoning, creative search formulation, and synthesis of contextual clues across time periods. Read our blog [here](https://parallel.ai/blog/deep-research)[here]($https://parallel.ai/blog/deep-research). 96 | 97 | ### ###Steps of reasoning 98 | 99 | 100% Multi-Hop questions 100 | 101 | ### ###Parallel 600/1200 102 | 103 | Parallel 600 and 1200 are agents with the same architecture as Parallel Ultra (which costs 300 USD for 1000 queries), but with 2x and 4x the compute and cost. 104 | 105 | Parallel-ultra establishes new state-of-the-art accuracy while remaining cost-efficient and our other processors complete the curve to establish the highest accuracy at each price point. This extends our track record from [SimpleQA and WISER-Atomic](https://parallel.ai/blog/parallel-task-api)[SimpleQA and WISER-Atomic]($https://parallel.ai/blog/parallel-task-api), demonstrating consistent leadership as research challenges scale from single-hop to complex multi-hop scenarios across a wide range of price points. 106 | 107 | OpenAI has published SOTA accuracy of 51.5% for their Deep Research Agent - trained on browse-comp tasks. This was achieved at an undisclosed computation shown on an exponential scale and isn’t available for API use. Since we’ve built our system to be able to optimize performance based on budgets for computation and retrieval, we were able to test our system at a budget level far beyond our Ultra processor with no changes to the underlying architecture. We observe (1) accuracy improves consistently with budget and (2) we were able to achieve 48% accuracy, without any optimization or fine-tuning on the dataset’s distribution. The implications extend beyond benchmarks: our customers can dial up performance for critical tasks or dial down performance for routine queries, providing flexibility unavailable in specialized systems. 108 | 109 | ##****Build with Parallel deep research**** 110 | ------------------------------------------- 111 | 112 | Get started building with the Parallel Task API pro and ultra processors in our Developer Platform or dive directly into our documentation. 113 | 114 | ###Run a Deep Research Query with the Parallel Task API 115 | 116 | 1 117 | 118 | 2 119 | 120 | 3 121 | 122 | 4 123 | 124 | 5 125 | 126 | 6 127 | 128 | 7 129 | 130 | 8 131 | 132 | 9 133 | 134 | 10 135 | 136 | 11 137 | 138 | 12 139 | 140 | 13 141 | 142 | 14 143 | 144 | ``` 145 | from parallel import Parallel 146 | 147 | # Initialize the Parallel client 148 | client = Parallel(api_key="your-api-key-here") 149 | 150 | # Execute the task run (blocking) 151 | run_result = client.task_run.execute( 152 | input="Company", 153 | output="Top adverse media, top risk factors, sample of customers,top competitors and their price/features/messaging", 154 | processor="ultra" 155 | ) 156 | print(run_result) 157 | ``` 158 | ```` from parallel import Parallel # Initialize the Parallel clientclient = Parallel(api_key="your-api-key-here") # Execute the task run (blocking)run_result = client.task_run.execute( input="Company", output="Top adverse media, top risk factors, sample of customers,top competitors and their price/features/messaging", processor="ultra")print(run_result)  ```` 159 | 160 | ##****Notes on Methodology**** 161 | ------------------------------ 162 | 163 | Benchmark Details: All benchmarks were run on a random 100 question subset of the original dataset, which was kept constant across experiments with our own agents and those of competitors. 164 | 165 | LLM Evaluator: The agents’ responses were compared against the ground truth using the same standard LLM evaluator and evaluation criteria. 166 | 167 | Benchmark Dates: All tests were conducted between Jun 10 and Jun 12, 2025. 168 | 169 | ![Image 3: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) 170 | 171 | By Parallel 172 | 173 | June 17, 2025 174 | 175 | ### ##Related Posts 10 176 | 177 | [![Image 4: A linear dithering of a search interface for agents](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fcf3d3b4112c8ab27af1a56442e549a6c9fb2b220-1600x1064.png&w=3840&q=75) ![Image 5: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[A state-of-the-art search API purpose-built for agents](https://parallel.ai/blog/search-api-benchmark) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/search-api-benchmark) 178 | 179 | [![Image 6: Parallel Search MCP Server in Devin](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fdfc00df5061085ec771f28d263b231d4c53747c3-1600x1044.png&w=3840&q=75) ![Image 7: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Parallel Search MCP Server in Devin](https://parallel.ai/blog/parallel-search-mcp-in-devin) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/parallel-search-mcp-in-devin) 180 | 181 | [![Image 8: Introducing Tool Calling via MCP Servers](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fe220ba8a89dfaeb943f943fc92636629a0b9ba14-1600x1600.png&w=3840&q=75) ![Image 9: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Tool Calling via MCP Servers](https://parallel.ai/blog/mcp-tool-calling) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/mcp-tool-calling) 182 | 183 | [![Image 10: Introducing the Parallel Search MCP Server ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F32cf3f1026d452dbaed58e8df4d7b487166d423a-1972x1972.png&w=3840&q=75) ![Image 11: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Search MCP Server ](https://parallel.ai/blog/search-mcp-server) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/search-mcp-server) 184 | 185 | [![Image 12: Starting today, Source Policy is available for both the Parallel Task API and Search API - giving you granular control over which sources your AI agents access and how results are prioritized.](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9eec7c7314218135df8fd7a83b1c6d3b953eece9-1528x1016.png&w=3840&q=75) ![Image 13: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Source Policy](https://parallel.ai/blog/source-policy) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/source-policy) 186 | 187 | [![Image 14: The Parallel Task Group API](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F5b63138a650248d4454f306ba72d99dc1572e08a-1600x946.png&w=3840&q=75) ![Image 15: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[The Parallel Task Group API](https://parallel.ai/blog/task-group-api) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/task-group-api) 188 | 189 | [![Image 16: Introducing the Parallel Search API ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Ff61c8461515390c7e546ee3ac1b470a4f79084cc-2560x1704.png&w=3840&q=75) ![Image 17: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Search API ](https://parallel.ai/blog/parallel-search-api) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/parallel-search-api) 190 | 191 | [![Image 18: Introducing the Parallel Chat API - a low latency web research API for web based LLM completions. The Parallel Chat API returns completions in text and structured JSON format, and is OpenAI Chat Completions compatible. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F780ef9d2c09f2e3526a0d11a4bda9eb0a45ef391-2560x1704.jpg&w=3840&q=75) ![Image 19: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Chat API ](https://parallel.ai/blog/chat-api) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/chat-api) 192 | 193 | [![Image 20: Parallel Web Systems introduces Basis with calibrated confidences - a new verification framework for AI web research and search API outputs that sets a new industry standard for transparent and reliable deep research. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F20273a28dbb0a5ee55784f4c9f7050427094f4eb-2672x1512.png&w=3840&q=75) ![Image 21: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Basis with Calibrated Confidences ](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) Tags:Product Release Reading time:4 min](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) 194 | 195 | [![Image 22: The Parallel Task API is a state-of-the-art system for automated web research that delivers the highest accuracy at every price point. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F96491c652ee4988fd9866d1b7619407582879e64-2576x1448.png&w=3840&q=75) ![Image 23: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Task API](https://parallel.ai/blog/parallel-task-api) Tags:Product Release,Benchmarks Reading time:4 min](https://parallel.ai/blog/parallel-task-api) 196 | 197 | ![Company Logo](https://parallel.ai/parallel-logo-540.png) 198 | 199 | ### Company 200 | 201 | * [hello@parallel.ai](mailto:hello@parallel.ai)[hello@parallel.ai](mailto:hello@parallel.ai) 202 | 203 | ### Resources 204 | 205 | * [About](https://parallel.ai/about)[About](https://parallel.ai/about) 206 | * [Docs](https://docs.parallel.ai/)[Docs](https://docs.parallel.ai) 207 | * [Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog) 208 | * [Changelog](https://docs.parallel.ai/resources/changelog)[Changelog](https://docs.parallel.ai/resources/changelog) 209 | * [Careers](https://jobs.ashbyhq.com/parallel)[Careers](https://jobs.ashbyhq.com/parallel) 210 | 211 | ### Info 212 | 213 | * [Terms](https://parallel.ai/terms-of-service)[Terms](https://parallel.ai/terms-of-service) 214 | * [Privacy](https://parallel.ai/privacy-policy)[Privacy](https://parallel.ai/privacy-policy) 215 | * [Trust Center](https://trust.parallel.ai/)[Trust Center](https://trust.parallel.ai/) 216 | 217 | ![SOC 2 Compliant](https://parallel.ai/soc2.svg) 218 | 219 | [LinkedIn](https://www.linkedin.com/company/parallel-web/about/)[LinkedIn](https://www.linkedin.com/company/parallel-web/about/) 220 | 221 | Parallel Web Systems Inc. 2025 222 | -------------------------------------------------------------------------------- /blog/introducing-basis-with-calibrated-confidences.md: -------------------------------------------------------------------------------- 1 | Title: Introducing Basis with Calibrated Confidences 2 | 3 | URL Source: https://parallel.ai/blog/introducing-basis-with-calibrated-confidences 4 | 5 | Published Time: 2025-05-16T20:26:10Z 6 | 7 | Markdown Content: 8 | A new standard for verifiable AI web research 9 | 10 | Tags:Product Release 11 | 12 | Reading time:4 min 13 | 14 | A month ago, we launched the Parallel Task API, powered by a series of processors that offer state of the art accuracy on web research tasks at every single price point. 15 | 16 | We recognize, however, that production use case don't just require pareto-optimal performance — they also require verifiability and calibrated confidence scoring and so today, we're excited to announce Basis — an essential suite of verification tools for the Parallel Task API. 17 | 18 | Basis is an automatically included add-on to our Core, Pro, and Ultra processors that provide additional context and evidence around how we came to our conclusion, as well as how confident we are in our findings. 19 | 20 | They allow you to identify instances where AI web research may yield unreliable results, enabling targeted human-in-the-loop workflows that efficiently focus human attention only where it's most needed. This strategic approach drastically reduces manual review hours while achieving significantly higher accuracy in hybrid human/AI workflows compared to either AI-only or human-only alternatives. 21 | 22 | ![Image 1: Confidence example as a competitor analysis task in Platform UI](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F0a5bf3b45a7c6fd1d26232847f72f6b182ae50b8-1280x852.png&w=3840&q=75) 23 | 24 | ![Confidence example as a competitor analysis task in Platform UI](https://cdn.sanity.io/images/5hzduz3y/production/0a5bf3b45a7c6fd1d26232847f72f6b182ae50b8-1280x852.png) 25 | 26 | ### ###**** 27 | 28 | What is Basis?**** 29 | 30 | Basis provides a complete framework for understanding and validating Task API outputs through four core components: 31 | 32 | * -****Citations****: Web URLs linking directly to source materials. 33 | * -****Reasoning****: Detailed explanations justifying each output field. 34 | * -****Excerpts****: Relevant text snippets from citation URLs. 35 | * -****Confidences****: A calibrated measure of confidence classified into low, medium, or high categories. 36 | 37 | These elements work together to create a robust framework for output verification that sets a new industry standard for transparency and reliability. For more information on Basis outputs in our Task API, go to [docs](https://docs.parallel.ai/core-concepts/basis)[docs]($https://docs.parallel.ai/core-concepts/basis). 38 | 39 | ### ###Calibrated Confidences 40 | 41 | Confidence ratings aren't particularly useful without calibration — you need to know that high, medium, and low labels provide useful and differentiable insight into task performance. 42 | 43 | To calibrate confidence ratings, we’ve tested confidence patterns on composite datasets that reflect a wide array of real world use cases. Each composite dataset has varying levels of difficulty to demonstrate confidence performance and distribution across any web research task. 44 | 45 | Why is this important? 46 | 47 | ##****Confidence provides insight into the difficulty of a Task**** 48 | ------------------------------------------------------------------- 49 | 50 | When Parallel returns a higher percentage of Basis outputs as "High" Confidence, you can reliably interpret this as Parallel's Task API performing well on the Task. 51 | 52 | You can use confidences as a proxy for a full evaluation, and understand how well your Tasks perform relative to each other. 53 | 54 | ##****Low confidence ratings provide efficient identification of what to review **** 55 | ------------------------------------------------------------------------------------ 56 | 57 | By reviewing just the outputs rated as "Low" Confidence—a small portion of your total dataset—you can typically achieve an ~2x reduction in error rate, giving you significantly more leverage over human time compared to reviewing all outputs. 58 | 59 | ##****High confidence ratings can be reliably skipped in manual review workflows**** 60 | ------------------------------------------------------------------------------------ 61 | 62 | Outputs marked as "High" Confidence have 2-3X lower error rates than that of the overall dataset. 63 | 64 | ### ###****Built for real-world applications at scale**** 65 | 66 | Basis is particularly valuable for hybrid AI-human workflows where the addition of AI significantly increases leverage, accuracy, and time efficiency. By focusing human review on outputs with low confidence, teams can dramatically reduce verification time while maintaining quality standards. This approach allows enterprises to scale their web research operations without sacrificing accuracy or transparency. 67 | 68 | Today, Basis powers human-in-the-loop production workflows across numerous domains. Insurance underwriters leverage low-confidence indicators and citation trails to streamline KYB verification processes that were previously manual. AI automation platforms use Basis to validate data enrichment capabilities before pushing to production, providing traceability from enriched fields back to source materials. 69 | 70 | The Basis framework with calibrated confidences is available today with the Parallel Task API. To start building with verifiable web research, go to the [Parallel Developer Platform](https://platform.parallel.ai/)[Parallel Developer Platform]($https://platform.parallel.ai/). 71 | 72 | ##****Notes on Methodology**** 73 | ------------------------------ 74 | 75 | ****Testing Dates****: Testing was conducted between May 12 and May 15, 2025 76 | 77 | ****Benchmark Details****: The Parallel Confidence datasets cover easy, medium, and hard web research questions that all reflect a wide range of representative real world use cases. Three example questions are below. 78 | 79 | ****Example Questions****: 80 | 81 | * -****Compliance**** 82 | 83 | Return if SOC2, ISO27001, PCI DSS are compliance frameworks mentioned on imsedge.com. If yes, return only the name of the compliance framework or frameworks that are mentioned. Otherwise, return no. 84 | * -****Company research**** 85 | 86 | Return if a company is B2B or B2C, the CEO’s Linkedin URL, the CEO’s name, the CEO’s undergrad institution, and the employee count of the company as of January 2025. 87 | * -****Financial research**** 88 | 89 | Find the SEC 10-K filing of the company as of January 2025 and the list of stock exchanges the company is listed on as of January 2025. 90 | 91 | ****Additional Data: **** 92 | 93 | ![Image 2: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) 94 | 95 | By Parallel 96 | 97 | May 16, 2025 98 | 99 | ### ##Related Posts 10 100 | -------------------------------------------------------------------------------- /blog/parallel-search-api.md: -------------------------------------------------------------------------------- 1 | Title: Introducing the Parallel Search API 2 | 3 | URL Source: https://parallel.ai/blog/parallel-search-api 4 | 5 | Published Time: 2025-06-10T20:03:07Z 6 | 7 | Markdown Content: 8 | Introducing the Parallel Search API | Enterprise Deep Research API for ChatGPT & Claude 9 | 10 | =============== 11 | 12 | [Parallel](https://parallel.ai/)[About](https://parallel.ai/about)[About](https://parallel.ai/about)[Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog)[Docs](https://docs.parallel.ai/introduction/quickstart)[Docs](https://docs.parallel.ai/introduction/quickstart) 13 | 14 | [Start Building P](https://platform.parallel.ai/)[Start Building](https://platform.parallel.ai/) 15 | 16 | Menu[Menu] 17 | 18 | Human Machine 19 | 20 | #Introducing the Parallel Search API 21 | ==================================== 22 | 23 | Tags:Product Release 24 | 25 | Reading time:2 min 26 | 27 | ![Image 1: Introducing the Parallel Search API ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Ff61c8461515390c7e546ee3ac1b470a4f79084cc-2560x1704.png&w=3840&q=75) 28 | 29 | Building AI agents and applications that access the web shouldn't require complex orchestration of searching, scraping, parsing, re-ranking, and filtering. The Parallel Search API handles this complexity for you, collapsing multi-step pipelines into a single fast API call. 30 | 31 | ##Why AIs Need a New Kind of Search 32 | ----------------------------------- 33 | 34 | LLMs ingest tokens, not web pages. Mainstream search engines are engineered for human use—short, keyword queries, clickable titles, and ad yield—so they surface teaser snippets instead of the high‑density passages an agent needs to reason. Developers are forced to add scraping and summarization layers that increase latency, inflate token costs, and introduce brittle failure points that can corrupt reliability and downstream quality. An AI‑native retrieval layer must deliver the most information‑rich spans of text, with explicit controls for freshness and length, ready to slot directly into an LLM context window. These requirements shape the Parallel Search API. 35 | 36 | ![Image 2: Search API Playground on the Parallel Developer Platform](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F7cb91842e91b65063cd8e35c4d3212309496a77c-1664x1080.gif&w=3840&q=75) 37 | 38 | ![Search API Playground on the Parallel Developer Platform](https://cdn.sanity.io/images/5hzduz3y/production/7cb91842e91b65063cd8e35c4d3212309496a77c-1664x1080.gif) 39 | 40 | Search API Playground on the Parallel Developer Platform 41 | 42 | ##****_\_ The \__ Web Search Tool for AI Agents**** 43 | --------------------------------------------------- 44 | 45 | Built on Parallel’s custom web crawler and index, the Search API takes flexible inputs (search objective and/or search queries) and returns LLM-ready ranked URLs with extended webpage excerpts. With granular control over output sizes, it largely reduces the need for additional scraping, making it the go-to search tool for your AI agent. 46 | 47 | ****Two tiers to match your needs:**** 48 | 49 | * -****Base****: Fast, cost-effective web access with extended webpage excerpts (2-5s) 50 | * -****Pro****: Best-in-class retrieval engine, prioritizing freshness and relevance. Built for long-horizon agents where quality matters over speed (15-60s) 51 | 52 | ###Create a Search API Request 53 | 54 | 1 55 | 56 | 2 57 | 58 | 3 59 | 60 | 4 61 | 62 | 5 63 | 64 | 6 65 | 66 | 7 67 | 68 | 8 69 | 70 | 9 71 | 72 | 10 73 | 74 | 11 75 | 76 | 12 77 | 78 | 13 79 | 80 | 14 81 | 82 | ``` 83 | curl --request POST \ 84 | --url https://api.parallel.ai/alpha/search \ 85 | --header "Content-Type: application/json" \ 86 | --header "x-api-key: $PARALLEL_API_KEY" \ 87 | --data '{ 88 | "objective": "When was the United Nations established? Prefer UN'\''s websites.", 89 | "search_queries": [ 90 | "Founding year UN", 91 | "Year of founding United Nations" 92 | ], 93 | "processor": "base", 94 | "max_results": 5, 95 | "max_chars_per_result": 1500 96 | }' 97 | ``` 98 | ```` curl --request POST \ --url https://api.parallel.ai/alpha/search \ --header "Content-Type: application/json" \ --header "x-api-key: $PARALLEL_API_KEY" \ --data '{ "objective": "When was the United Nations established? Prefer UN'\''s websites.", "search_queries": [ "Founding year UN", "Year of founding United Nations" ], "processor": "base", "max_results": 5, "max_chars_per_result": 1500 }'```` 99 | 100 | The Parallel Search API delivers high-quality, relevant results while optimizing for the price-performance balance your AI applications need at scale. By providing a single, simple abstraction, our Search API reduces token spend and eliminates the need to orchestrate multiple tools. Our [Chat](https://parallel.ai/blog/chat-api)[Chat]($https://parallel.ai/blog/chat-api) and [Task APIs](https://parallel.ai/blog/parallel-task-api)[Task APIs]($https://parallel.ai/blog/parallel-task-api) utilize this same search technology as their underlying foundation. 101 | 102 | ##****Start Building**** 103 | ------------------------ 104 | 105 | Get started in our [Developer Platform](https://platform.parallel.ai/play/search)[Developer Platform]($https://platform.parallel.ai/play/search) or dive into the [documentation](https://docs.parallel.ai/resources/search-api)[documentation]($https://docs.parallel.ai/resources/search-api). 106 | 107 | ![Image 3: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) 108 | 109 | By Parallel 110 | 111 | June 10, 2025 112 | 113 | ### ##Related Posts 10 114 | 115 | [![Image 4: A linear dithering of a search interface for agents](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fcf3d3b4112c8ab27af1a56442e549a6c9fb2b220-1600x1064.png&w=3840&q=75) ![Image 5: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[A state-of-the-art search API purpose-built for agents](https://parallel.ai/blog/search-api-benchmark) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/search-api-benchmark) 116 | 117 | [![Image 6: Parallel Search MCP Server in Devin](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fdfc00df5061085ec771f28d263b231d4c53747c3-1600x1044.png&w=3840&q=75) ![Image 7: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Parallel Search MCP Server in Devin](https://parallel.ai/blog/parallel-search-mcp-in-devin) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/parallel-search-mcp-in-devin) 118 | 119 | [![Image 8: Introducing Tool Calling via MCP Servers](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fe220ba8a89dfaeb943f943fc92636629a0b9ba14-1600x1600.png&w=3840&q=75) ![Image 9: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Tool Calling via MCP Servers](https://parallel.ai/blog/mcp-tool-calling) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/mcp-tool-calling) 120 | 121 | [![Image 10: Introducing the Parallel Search MCP Server ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F32cf3f1026d452dbaed58e8df4d7b487166d423a-1972x1972.png&w=3840&q=75) ![Image 11: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Search MCP Server ](https://parallel.ai/blog/search-mcp-server) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/search-mcp-server) 122 | 123 | [![Image 12: Starting today, Source Policy is available for both the Parallel Task API and Search API - giving you granular control over which sources your AI agents access and how results are prioritized.](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9eec7c7314218135df8fd7a83b1c6d3b953eece9-1528x1016.png&w=3840&q=75) ![Image 13: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Source Policy](https://parallel.ai/blog/source-policy) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/source-policy) 124 | 125 | [![Image 14: The Parallel Task Group API](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F5b63138a650248d4454f306ba72d99dc1572e08a-1600x946.png&w=3840&q=75) ![Image 15: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[The Parallel Task Group API](https://parallel.ai/blog/task-group-api) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/task-group-api) 126 | 127 | [![Image 16: State of the Art Deep Research APIs](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F897a6d4a4d5caf5657b59cdc1dc99de0641c66df-1600x900.png&w=3840&q=75) ![Image 17: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[State of the Art Deep Research APIs](https://parallel.ai/blog/deep-research) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/deep-research) 128 | 129 | [![Image 18: Introducing the Parallel Chat API - a low latency web research API for web based LLM completions. The Parallel Chat API returns completions in text and structured JSON format, and is OpenAI Chat Completions compatible. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F780ef9d2c09f2e3526a0d11a4bda9eb0a45ef391-2560x1704.jpg&w=3840&q=75) ![Image 19: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Chat API ](https://parallel.ai/blog/chat-api) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/chat-api) 130 | 131 | [![Image 20: Parallel Web Systems introduces Basis with calibrated confidences - a new verification framework for AI web research and search API outputs that sets a new industry standard for transparent and reliable deep research. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F20273a28dbb0a5ee55784f4c9f7050427094f4eb-2672x1512.png&w=3840&q=75) ![Image 21: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Basis with Calibrated Confidences ](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) Tags:Product Release Reading time:4 min](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) 132 | 133 | [![Image 22: The Parallel Task API is a state-of-the-art system for automated web research that delivers the highest accuracy at every price point. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F96491c652ee4988fd9866d1b7619407582879e64-2576x1448.png&w=3840&q=75) ![Image 23: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Task API](https://parallel.ai/blog/parallel-task-api) Tags:Product Release,Benchmarks Reading time:4 min](https://parallel.ai/blog/parallel-task-api) 134 | 135 | ![Company Logo](https://parallel.ai/parallel-logo-540.png) 136 | 137 | ### Company 138 | 139 | * [hello@parallel.ai](mailto:hello@parallel.ai)[hello@parallel.ai](mailto:hello@parallel.ai) 140 | 141 | ### Resources 142 | 143 | * [About](https://parallel.ai/about)[About](https://parallel.ai/about) 144 | * [Docs](https://docs.parallel.ai/)[Docs](https://docs.parallel.ai) 145 | * [Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog) 146 | * [Changelog](https://docs.parallel.ai/resources/changelog)[Changelog](https://docs.parallel.ai/resources/changelog) 147 | * [Careers](https://jobs.ashbyhq.com/parallel)[Careers](https://jobs.ashbyhq.com/parallel) 148 | 149 | ### Info 150 | 151 | * [Terms](https://parallel.ai/terms-of-service)[Terms](https://parallel.ai/terms-of-service) 152 | * [Privacy](https://parallel.ai/privacy-policy)[Privacy](https://parallel.ai/privacy-policy) 153 | * [Trust Center](https://trust.parallel.ai/)[Trust Center](https://trust.parallel.ai/) 154 | 155 | ![SOC 2 Compliant](https://parallel.ai/soc2.svg) 156 | 157 | [LinkedIn](https://www.linkedin.com/company/parallel-web/about/)[LinkedIn](https://www.linkedin.com/company/parallel-web/about/) 158 | 159 | Parallel Web Systems Inc. 2025 160 | -------------------------------------------------------------------------------- /blog/parallel-search-mcp-in-devin.md: -------------------------------------------------------------------------------- 1 | Title: Parallel Search MCP Server in Devin 2 | 3 | URL Source: https://parallel.ai/blog/parallel-search-mcp-in-devin 4 | 5 | Published Time: 2025-07-31T04:55:05Z 6 | 7 | Markdown Content: 8 | Parallel Search MCP Server in Devin | Enterprise Deep Research API for ChatGPT & Claude 9 | 10 | =============== 11 | 12 | [Parallel](https://parallel.ai/)[About](https://parallel.ai/about)[About](https://parallel.ai/about)[Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog)[Docs](https://docs.parallel.ai/introduction/quickstart)[Docs](https://docs.parallel.ai/introduction/quickstart) 13 | 14 | [Start Building P](https://platform.parallel.ai/)[Start Building](https://platform.parallel.ai/) 15 | 16 | Menu[Menu] 17 | 18 | Human Machine 19 | 20 | #Parallel Search MCP Server in Devin 21 | ==================================== 22 | 23 | Tags:Product Release 24 | 25 | Reading time:2 min 26 | 27 | ![Image 1: Parallel Search MCP Server in Devin](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fdfc00df5061085ec771f28d263b231d4c53747c3-1600x1044.png&w=3840&q=75) 28 | 29 | The [Parallel Search MCP Server](https://parallel.ai/blog/search-mcp-server)[Parallel Search MCP Server]($https://parallel.ai/blog/search-mcp-server) is now live in [Devin's MCP Marketplace](https://docs.devin.ai/work-with-devin/mcp)[Devin's MCP Marketplace]($https://docs.devin.ai/work-with-devin/mcp), bringing high quality web research capabilities directly to the AI software engineer. [Devin](https://devin.ai/)[Devin]($https://devin.ai/) can now query the web with the same infrastructure that powers our state-of-the-art [Task](https://parallel.ai/blog/parallel-task-api)[Task]($https://parallel.ai/blog/parallel-task-api) and [Chat APIs](https://parallel.ai/blog/chat-api)[Chat APIs]($https://parallel.ai/blog/chat-api). 30 | 31 | ##****Web-Aware Development at Scale**** 32 | ---------------------------------------- 33 | 34 | With Parallel Search MCP enabled, Devin accesses ranked URLs with information-rich excerpts optimized for AI reasoning - no complex orchestration of scraping, parsing, and filtering required. This transforms how Devin approaches development workflows that demand real-time web intelligence. 35 | 36 | ****Live debugging and troubleshooting****: When encountering errors, Devin queries Stack Overflow, GitHub issues, and technical forums for solutions, accessing the collective knowledge of the developer community in real-time. 37 | 38 | ****Learning from online codebases****: Browse open-source projects on GitHub to see how experienced developers solve similar problems and adopt proven patterns. 39 | 40 | ****API research****: Devin retrieves the latest API documentation, framework guides, and technical specifications while coding, ensuring it works with current information rather than potentially outdated training data. 41 | 42 | ****Package and dependency intelligence****:********Check for the latest library versions, read changelogs, and understand compatibility requirements in real-time, making dependency management more reliable and informed. 43 | 44 | [![Image 2: Illustration demonstrating deep research API concepts, web search capabilities, or AI agent integration features](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fde636b7d3051229eb711d95ec46b63944143dba1-1316x1080.gif&w=3840&q=75)](https://docs.devin.ai/work-with-devin/mcp) 45 | 46 | ![](https://cdn.sanity.io/images/5hzduz3y/production/de636b7d3051229eb711d95ec46b63944143dba1-1316x1080.gif) 47 | 48 | Parallel Search MCP in Devin 49 | 50 | ##****Get Started**** 51 | --------------------- 52 | 53 | We’re excited to see what you build with a web-aware Devin. Get started by visiting the [Devin MCP Marketplace](https://docs.devin.ai/work-with-devin/mcp)[Devin MCP Marketplace]($https://docs.devin.ai/work-with-devin/mcp). 54 | 55 | ![Image 3: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) 56 | 57 | By Parallel 58 | 59 | July 31, 2025 60 | 61 | ### ##Related Posts 10 62 | 63 | [![Image 4: A linear dithering of a search interface for agents](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fcf3d3b4112c8ab27af1a56442e549a6c9fb2b220-1600x1064.png&w=3840&q=75) ![Image 5: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[A state-of-the-art search API purpose-built for agents](https://parallel.ai/blog/search-api-benchmark) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/search-api-benchmark) 64 | 65 | [![Image 6: Introducing Tool Calling via MCP Servers](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fe220ba8a89dfaeb943f943fc92636629a0b9ba14-1600x1600.png&w=3840&q=75) ![Image 7: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Tool Calling via MCP Servers](https://parallel.ai/blog/mcp-tool-calling) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/mcp-tool-calling) 66 | 67 | [![Image 8: Introducing the Parallel Search MCP Server ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F32cf3f1026d452dbaed58e8df4d7b487166d423a-1972x1972.png&w=3840&q=75) ![Image 9: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Search MCP Server ](https://parallel.ai/blog/search-mcp-server) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/search-mcp-server) 68 | 69 | [![Image 10: Starting today, Source Policy is available for both the Parallel Task API and Search API - giving you granular control over which sources your AI agents access and how results are prioritized.](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9eec7c7314218135df8fd7a83b1c6d3b953eece9-1528x1016.png&w=3840&q=75) ![Image 11: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Source Policy](https://parallel.ai/blog/source-policy) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/source-policy) 70 | 71 | [![Image 12: The Parallel Task Group API](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F5b63138a650248d4454f306ba72d99dc1572e08a-1600x946.png&w=3840&q=75) ![Image 13: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[The Parallel Task Group API](https://parallel.ai/blog/task-group-api) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/task-group-api) 72 | 73 | [![Image 14: State of the Art Deep Research APIs](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F897a6d4a4d5caf5657b59cdc1dc99de0641c66df-1600x900.png&w=3840&q=75) ![Image 15: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[State of the Art Deep Research APIs](https://parallel.ai/blog/deep-research) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/deep-research) 74 | 75 | [![Image 16: Introducing the Parallel Search API ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Ff61c8461515390c7e546ee3ac1b470a4f79084cc-2560x1704.png&w=3840&q=75) ![Image 17: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Search API ](https://parallel.ai/blog/parallel-search-api) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/parallel-search-api) 76 | 77 | [![Image 18: Introducing the Parallel Chat API - a low latency web research API for web based LLM completions. The Parallel Chat API returns completions in text and structured JSON format, and is OpenAI Chat Completions compatible. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F780ef9d2c09f2e3526a0d11a4bda9eb0a45ef391-2560x1704.jpg&w=3840&q=75) ![Image 19: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Chat API ](https://parallel.ai/blog/chat-api) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/chat-api) 78 | 79 | [![Image 20: Parallel Web Systems introduces Basis with calibrated confidences - a new verification framework for AI web research and search API outputs that sets a new industry standard for transparent and reliable deep research. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F20273a28dbb0a5ee55784f4c9f7050427094f4eb-2672x1512.png&w=3840&q=75) ![Image 21: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Basis with Calibrated Confidences ](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) Tags:Product Release Reading time:4 min](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) 80 | 81 | [![Image 22: The Parallel Task API is a state-of-the-art system for automated web research that delivers the highest accuracy at every price point. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F96491c652ee4988fd9866d1b7619407582879e64-2576x1448.png&w=3840&q=75) ![Image 23: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Task API](https://parallel.ai/blog/parallel-task-api) Tags:Product Release,Benchmarks Reading time:4 min](https://parallel.ai/blog/parallel-task-api) 82 | 83 | ![Company Logo](https://parallel.ai/parallel-logo-540.png) 84 | 85 | ### Company 86 | 87 | * [hello@parallel.ai](mailto:hello@parallel.ai)[hello@parallel.ai](mailto:hello@parallel.ai) 88 | 89 | ### Resources 90 | 91 | * [About](https://parallel.ai/about)[About](https://parallel.ai/about) 92 | * [Docs](https://docs.parallel.ai/)[Docs](https://docs.parallel.ai) 93 | * [Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog) 94 | * [Changelog](https://docs.parallel.ai/resources/changelog)[Changelog](https://docs.parallel.ai/resources/changelog) 95 | * [Careers](https://jobs.ashbyhq.com/parallel)[Careers](https://jobs.ashbyhq.com/parallel) 96 | 97 | ### Info 98 | 99 | * [Terms](https://parallel.ai/terms-of-service)[Terms](https://parallel.ai/terms-of-service) 100 | * [Privacy](https://parallel.ai/privacy-policy)[Privacy](https://parallel.ai/privacy-policy) 101 | * [Trust Center](https://trust.parallel.ai/)[Trust Center](https://trust.parallel.ai/) 102 | 103 | ![SOC 2 Compliant](https://parallel.ai/soc2.svg) 104 | 105 | [LinkedIn](https://www.linkedin.com/company/parallel-web/about/)[LinkedIn](https://www.linkedin.com/company/parallel-web/about/) 106 | 107 | Parallel Web Systems Inc. 2025 108 | -------------------------------------------------------------------------------- /blog/parallel-task-api.md: -------------------------------------------------------------------------------- 1 | Title: Introducing the Parallel Task API 2 | 3 | URL Source: https://parallel.ai/blog/parallel-task-api 4 | 5 | Published Time: 2025-04-24T21:32:51Z 6 | 7 | Markdown Content: 8 | The state-of-the-art system for automated web research 9 | 10 | Tags:Product Release,Benchmarks 11 | 12 | Reading time:4 min 13 | 14 | At [Parallel Web Systems](https://parallel.ai/)[Parallel Web Systems]($https://parallel.ai/), we’ve been busy [building infrastructure for a web built for AIs](https://parallel.ai/about)[building infrastructure for a web built for AIs]($/about). Today, we’re excited to announce our first product: the Parallel Task API. 15 | 16 | The [Parallel Task API](https://docs.parallel.ai/)[Parallel Task API]($https://docs.parallel.ai) offers developers and enterprises a robust, reliable, and highly scalable solution for obtaining data, research findings, and insights from the internet. It has been built and optimized for repeated workflows over information on the open web that demand scalability and quality. Instead of engineering bespoke and complex AI+human workflows, users simply declare what information and insights they need. Our API seamlessly orchestrates the querying, ranking, retrieval, reasoning, validation, and synthesis to deliver this information as structured outputs. 17 | 18 | Beyond simplifying the interface for our customers, our declarative approach enables optimization across a wide range of query plans, resulting in unprecedented accuracy. We provide customers with a choice of processors, each with a budget across compute and retrieval, and automatically optimized to achieve peak performance. This allows us to offer simple, transparent, and predictable pricing while delivering the highest quality results across a wide range of use cases across a wide range of price points. 19 | 20 | Every response includes comprehensive citations linking to source materials and detailed reasoning for every output field. For most processors, we also provide confidence scores that are calibrated to reflect the uncertainty for each output field. These features make the Parallel Task API an ideal choice for production systems and workflows with the most exacting requirements for accuracy, scale, and auditability. 21 | 22 | ##State-of-the-art performance, priced for scaled use 23 | ----------------------------------------------------- 24 | 25 | The Parallel Task API sets a new standard on web research benchmarks, both academic and ones we’ve built inspired by actual customer use cases spanning several domains. 26 | 27 | While [SimpleQA](https://openai.com/index/introducing-simpleqa/)[SimpleQA]($https://openai.com/index/introducing-simpleqa/) measures factual accuracy on straightforward questions, we developed a proprietary benchmark suite called WISER that tests real-world, customer-inspired use cases across financial research, sales intelligence, recruitment, and other domains. Beyond just achieving the highest accuracy score, we outperform leading alternatives at every price point establishing a new pareto-frontier for both SimpleQA and WISER-Atomic. 28 | 29 | ##Motivated and designed for real-world applications, built for scale 30 | --------------------------------------------------------------------- 31 | 32 | Today, Parallel powers production-grade workflows across a range of domains. As a SOC-II Type 2 certified platform, we're trusted by startups and enterprises alike to get structured insights from the web reliably and at scale. Our API serves diverse use cases across domains: insurance underwriters evaluating business risks, sales teams enriching CRM data with company intelligence, financial analysts conducting due diligence or sourcing, and professional service firms streamlining manual research workflows. The Task API is available today - we are excited to discover and serve use cases beyond our imagination. 33 | 34 | ![Image 1: Illustration demonstrating deep research API concepts, web search capabilities, or AI agent integration features](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F92663b0f1f8fa316d21743f04d7c32d9b7404d00-2576x1712.png&w=3840&q=75) 35 | 36 | ![](https://cdn.sanity.io/images/5hzduz3y/production/92663b0f1f8fa316d21743f04d7c32d9b7404d00-2576x1712.png) 37 | 38 | To start building on our developer platform, [request access](https://parallel.ai/)[request access]($https://parallel.ai). 39 | 40 | ##_\_ Notes on methodology \__ 41 | ------------------------------ 42 | 43 | ****Standalone LLMs:**** While some might be tempted to compare Parallel to plain LLMs (e.g., GPT-4.5), we consider this a different category because standalone LLMs lack real-time web access. That said, the best standalone model we tested—GPT-4.5—achieved around 62.5% SimpleQA accuracy, which is impressive in its own right. However, it cannot update from current data or retrieve verified web sources in real time. 44 | 45 | ****Models and competitors used: ****For providers with publicly comparable AI-based web search APIs, we used all available configurations. However, given some providers have a large number of configurations, we show and report only the configurations on the provider’s own price-performance frontier on the WISER-Atomic benchmark. For SimpleQA, we chose 2-3 diverse configurations per provider to measure and report.********Perplexity reports an accuracy score of 93.9% for their Deep Research implementation on SimpleQA. On WISER-Atomic, Perplexity Deep Research scores 64% with an average cost of 290 CPM over the benchmark. 46 | 47 | ****LLM Evaluator:**** Each system's responses were scored by the same standard with the same evaluator and evaluation criteria. For SimpleQA, we used the standard evaluator. For our internal benchmark, we used our proprietary evaluator. 48 | 49 | ****Benchmark Dates:**** Testing was conducted between April 21 and April 24, 2025. 50 | 51 | ****Benchmark Details: ****For all models, we ran the full SimpleQA benchmark (~4,300 questions in total). While our internal benchmark is large and contains more complex tasks, we report on a test set of 121 questions that we refer to as the WISER-Atomic (Web Intelligence Search Evaluation Research) dataset. Three example questions are below. 52 | 53 | ****Limitations: ****There are differences in the latency of each model, which we do not report in this post. Most notably, the Parallel processors tested are slower than the competitors. 54 | 55 | ****Cost Standardization:**** Pricing models vary across all providers benchmarked. Parallel has a standardized and transparent per-query pricing model. For APIs with token-based billing (OpenAI, Perplexity, etc.), we used publicly listed prices and normalized them to a consistent cost per thousand queries (CPM) as measured on the benchmark. As a result, the prices for the same model may vary across benchmarks. 56 | 57 | ****Example questions from WISER-Atomic**** 58 | 59 | ![Image 2: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) 60 | 61 | By Parallel 62 | 63 | April 24, 2025 64 | 65 | ### ##Related Posts 10 66 | -------------------------------------------------------------------------------- /blog/search-api-benchmark.md: -------------------------------------------------------------------------------- 1 | Title: A state-of-the-art search API purpose-built for agents 2 | 3 | URL Source: https://parallel.ai/blog/search-api-benchmark 4 | 5 | Published Time: 2025-07-31T22:14:17Z 6 | 7 | Markdown Content: 8 | The Parallel Search MCP Server offers an easy to integrate, state-of-the-art, web search solution for AI agents. Built on the same search infrastructure that powers Parallel’s Task API and Search API, it demonstrates superior performance while being up to 50% cheaper than LLM-native web search implementations - establishing a new price-performance frontier for AI agent web access. 9 | 10 | Tags:Benchmarks 11 | 12 | Reading time:3 min 13 | 14 | ##****Rethinking web search for AI agents**** 15 | --------------------------------------------- 16 | 17 | Mainstream search engines are designed for human browsing patterns - keyword queries, short snippets designed to drive clicks, and ad-optimized layouts - rather than the information-dense passages AI agents need to reason effectively. 18 | 19 | When building our higher-level [Task API](https://parallel.ai/blog/parallel-task-api)[Task API]($https://parallel.ai/blog/parallel-task-api), we recognized this mismatch early and built our own Search API purpose-built for AI agents. The Parallel Search API accepts broader declarative task objectives beyond simple keyword queries, allowing for more complex searches. It also manages agent context by returning the most relevant dense excerpts in an LLM-friendly format, instead of incomplete snippets or full-page text. For agentic pipelines, this translates to fewer input tokens (reduced cost), better signal-to-noise to reason over (improved quality), and research that concludes in fewer steps (lower end-to-end latency). 20 | 21 | ****The result:**** a simple one-shot interface for agent web access. This replaces multi-step search/scrape/extract/rerank pipelines that increase latency, inflate token costs, and introduce failure points that break agent workflows. 22 | 23 | ##****Leading performance at the lowest cost**** 24 | ------------------------------------------------ 25 | 26 | To evaluate real-world performance of the Parallel Search MCP Server, we created the WISER-Search benchmark which blends WISER-Fresh (queries requiring the freshest data from the web) and WISER-Atomic (hard real-world business queries). This combination reflects the challenges AI agents face in production environments across breaking news, financial data, technical documentation, and competitive intelligence. 27 | 28 | Sample questions include: 29 | 30 | ### WISER-Fresh 31 | 32 | * -Which automaker signed today’s major chip‑supply deal with Samsung Electronics? 33 | * -Which HR software firm did EQT agree to buy today? 34 | * -How many shares does Firefly Aerospace plan to offer in its IPO filing? 35 | * -Which Azerbaijani energy firm signed Ukraine’s first Transbalkan gas deal? 36 | * -What revenue range did Audi forecast after cutting its guidance in July 2025? 37 | 38 | ### WISER-Atomic 39 | 40 | * -In fiscal year 2024, what percentage of Salesforce's subscription and support revenue came from the segment that includes its Tableau acquisition, and how does this compare to the company's overall CRM market share in 2023? Please share all of your factual findings that helped you answer the question, in your final answer. 41 | * -According to the International Debt Statistics 2023 by the World Bank, calculate the average Foreign Direct Investment amount (in millions of USD) for Sri Lanka,Turkmenistan, and Niger in 2019. Round your answer to two decimal places. 42 | * -Navigate to the website http://www.flightaware.com. This is the main domain for the target company. Once you are on their website, locate their careers or jobs page. Print the careers page URL. 43 | 44 | Results on the blended WISER-Search benchmark, comparing three different web search solutions (Parallel MCP server, Exa MCP server/tool calling, native web search) across four different LLMs (GPT 4.1, O4-mini, O3, Claude Sonnet 4), are shown below. 45 | 46 | ****The results show that agents using Parallel Search MCP achieve superior accuracy at up to 50% lower total cost**** when compared to agents using native web search implementations. Agentic workflows using the Parallel Search MCP conduct fewer tool calls and receive denser excerpts to reason on. As a result, the total cost (Search API cost + LLM cost) and latency are meaningfully reduced, while producing higher quality results. 47 | 48 | ##****Easily replace LLM native search with Parallel Search MCP**** 49 | ------------------------------------------------------------------- 50 | 51 | If you're building an AI agent that needs web access, the Parallel Search MCP Server is easy to integrate with any MCP-aware LLM. Simply change one parameter and see immediate results. 52 | 53 | [![Image 1: Example of replacing OpenAI Native Search with Parallel Search MCP](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fd0cf62816792b6e5b5d39c923518df9b87a64fa6-2414x1555.png&w=3840&q=75)](https://parallel.ai/blog/search-api-benchmark) 54 | 55 | ![Example of replacing OpenAI Native Search with Parallel Search MCP](https://cdn.sanity.io/images/5hzduz3y/production/d0cf62816792b6e5b5d39c923518df9b87a64fa6-2414x1555.png) 56 | 57 | Start building with state-of-the-art web search purpose-built for agents today. Get started in our [Developer Platform](https://platform.parallel.ai/play/search)[Developer Platform]($https://platform.parallel.ai/play/search) or dive directly into [Documentation](https://docs.parallel.ai/features/remote-mcp)[Documentation]($https://docs.parallel.ai/features/remote-mcp). 58 | 59 | ##****Methodology**** 60 | --------------------- 61 | 62 | ****Benchmark details****: All tests were conducted on a dataset spanning real-world scenarios including breaking news, financial data, technical documentation, and competitive intelligence queries. The dataset is a combination of WISER-Fresh (76 easily verifiable questions based on events on a current day, generated by OpenAI o3 pro) and WISER-Atomic (120 questions based on real world use cases from Parallel customers). 63 | 64 | ****Evaluation****: Responses were evaluated using standardized LLM evaluators measuring accuracy against verified ground truth answers. 65 | 66 | ****Cost calculation****: Cost reflects the average cost per query across all questions run. This cost includes both the search API call and LLM token cost. 67 | 68 | ****Testing dates****: WISER-Fresh data was generated on July 28th, 2025 and testing was conducted within 24 hrs of dataset generation. WISER-Atomic testing was conducted from July 28th, 2025 to July 29th, 2025. 69 | 70 | ![Image 2: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) 71 | 72 | By Parallel 73 | 74 | July 31, 2025 75 | 76 | ### ##Related Posts 10 77 | -------------------------------------------------------------------------------- /blog/search-mcp-server.md: -------------------------------------------------------------------------------- 1 | Title: Introducing the Parallel Search MCP Server 2 | 3 | URL Source: https://parallel.ai/blog/search-mcp-server 4 | 5 | Published Time: 2025-07-14T23:20:18Z 6 | 7 | Markdown Content: 8 | Introducing the Parallel Search MCP Server | Enterprise Deep Research API for ChatGPT & Claude 9 | 10 | =============== 11 | 12 | [Parallel](https://parallel.ai/)[About](https://parallel.ai/about)[About](https://parallel.ai/about)[Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog)[Docs](https://docs.parallel.ai/introduction/quickstart)[Docs](https://docs.parallel.ai/introduction/quickstart) 13 | 14 | [Start Building P](https://platform.parallel.ai/)[Start Building](https://platform.parallel.ai/) 15 | 16 | Menu[Menu] 17 | 18 | Human Machine 19 | 20 | #Introducing the Parallel Search MCP Server 21 | =========================================== 22 | 23 | Tags:Product Release 24 | 25 | Reading time:2 min 26 | 27 | ![Image 1: Introducing the Parallel Search MCP Server ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F32cf3f1026d452dbaed58e8df4d7b487166d423a-1972x1972.png&w=3840&q=75) 28 | 29 | Last month, we unveiled the [Parallel Search API](https://parallel.ai/blog/parallel-search-api)[Parallel Search API]($https://parallel.ai/blog/parallel-search-api) - a single endpoint purpose-built for AIs that takes in flexible search objectives and outputs high-density, LLM-ready search results with extended snippets. Today, we’re making that same capability available out of the box for any model that supports tool use, via the Parallel Search MCP Server. 30 | 31 | The MCP Server exposes our Search API as a plug-and-play tool, giving LLMs instant access to real-time web knowledge with a simple configuration change. This replaces brittle, multi-step search stacks with a single, production-ready tool that delivers higher quality results at significantly lower cost. 32 | 33 | [![Image 2: Illustration demonstrating deep research API concepts, web search capabilities, or AI agent integration features](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F520765597d48204a4b468d911ca4dc1fb9d617bb-1708x1080.gif&w=3840&q=75)](https://parallel.ai/blog/mcp-server-in-claude) 34 | 35 | ![](https://cdn.sanity.io/images/5hzduz3y/production/520765597d48204a4b468d911ca4dc1fb9d617bb-1708x1080.gif) 36 | 37 | The Parallel Search MCP server in Claude 38 | 39 | ##****MCP-Native Search**** 40 | --------------------------- 41 | 42 | The Search MCP server allows developers to easily integrate the Parallel Search API with any MCP-aware LLM, eliminating the complexity of custom API wrappers. The MCP Server delivers: 43 | 44 | * -****Seamless integration****: Plug-and-play with OpenAI, Anthropic, and other MCP-aware clients - no custom REST wiring. 45 | * -****Flexible search inputs****: Works with both natural language objectives and keyword queries, with precise controls for domains, freshness, and length. 46 | * -****Superior quality at lower cost****: Dense, citation-rich passages ranked for LLM reasoning, delivered for a fraction of typical search spend. 47 | * -****Production scale****: Built on Parallel's proprietary web crawler and index, the same infrastructure that powers our public Search API at scale. 48 | 49 | ##****A Quick Integration**** 50 | ----------------------------- 51 | 52 | Adding Parallel Search to your LLM client is as simple as swapping in a simple tool definition: 53 | 54 | ###Add Parallel to your LLM client 55 | 56 | 1 57 | 58 | 2 59 | 60 | 3 61 | 62 | 4 63 | 64 | 5 65 | 66 | 6 67 | 68 | 7 69 | 70 | 8 71 | 72 | 9 73 | 74 | 10 75 | 76 | 11 77 | 78 | 12 79 | 80 | 13 81 | 82 | 14 83 | 84 | 15 85 | 86 | 16 87 | 88 | 17 89 | 90 | 18 91 | 92 | 19 93 | 94 | 20 95 | 96 | 21 97 | 98 | 22 99 | 100 | 23 101 | 102 | 24 103 | 104 | 25 105 | 106 | 26 107 | 108 | 27 109 | 110 | ``` 111 | from openai import OpenAI 112 | from openai.types import responses as openai_responses 113 | 114 | PARALLEL_API_KEY = "your-api-key" 115 | 116 | tools = [ 117 | openai_responses.tool_param.Mcp( 118 | server_label="parallel_web_search", 119 | server_url="https://mcp.parallel.ai/alpha/search_mcp/", 120 | headers={"x-api-key": PARALLEL_API_KEY}, 121 | type="mcp", 122 | require_approval="never", 123 | ) 124 | ] 125 | 126 | client = OpenAI() 127 | resp = client.responses.create( 128 | model="gpt-4o-mini", 129 | input="What are the latest findings in fusion energy?", 130 | tools=tools, 131 | tool_choice="required", 132 | ) 133 | print(resp) 134 | ``` 135 | ```` from openai import OpenAIfrom openai.types import responses as openai_responses  PARALLEL_API_KEY = "your-api-key"  tools = [ openai_responses.tool_param.Mcp( server_label="parallel_web_search", server_url="https://mcp.parallel.ai/alpha/search_mcp/", headers={"x-api-key": PARALLEL_API_KEY}, type="mcp", require_approval="never", )]  client = OpenAI()resp = client.responses.create( model="gpt-4o-mini", input="What are the latest findings in fusion energy?", tools=tools, tool_choice="required",)print(resp) ```` 136 | 137 | ##****Start Building**** 138 | ------------------------ 139 | 140 | Connect to the Parallel Search MCP Server through your MCP-compatible client and start accessing real-time web knowledge instantly. Get started in our [Developer Platform](https://platform.parallel.ai/)[Developer Platform]($https://platform.parallel.ai/) or dive into the [documentation](https://docs.parallel.ai/features/remote-mcp)[documentation]($https://docs.parallel.ai/features/remote-mcp). 141 | 142 | ![Image 3: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) 143 | 144 | By Parallel 145 | 146 | July 14, 2025 147 | 148 | ### ##Related Posts 10 149 | 150 | [![Image 4: A linear dithering of a search interface for agents](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fcf3d3b4112c8ab27af1a56442e549a6c9fb2b220-1600x1064.png&w=3840&q=75) ![Image 5: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[A state-of-the-art search API purpose-built for agents](https://parallel.ai/blog/search-api-benchmark) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/search-api-benchmark) 151 | 152 | [![Image 6: Parallel Search MCP Server in Devin](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fdfc00df5061085ec771f28d263b231d4c53747c3-1600x1044.png&w=3840&q=75) ![Image 7: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Parallel Search MCP Server in Devin](https://parallel.ai/blog/parallel-search-mcp-in-devin) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/parallel-search-mcp-in-devin) 153 | 154 | [![Image 8: Introducing Tool Calling via MCP Servers](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fe220ba8a89dfaeb943f943fc92636629a0b9ba14-1600x1600.png&w=3840&q=75) ![Image 9: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Tool Calling via MCP Servers](https://parallel.ai/blog/mcp-tool-calling) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/mcp-tool-calling) 155 | 156 | [![Image 10: Starting today, Source Policy is available for both the Parallel Task API and Search API - giving you granular control over which sources your AI agents access and how results are prioritized.](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9eec7c7314218135df8fd7a83b1c6d3b953eece9-1528x1016.png&w=3840&q=75) ![Image 11: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Source Policy](https://parallel.ai/blog/source-policy) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/source-policy) 157 | 158 | [![Image 12: The Parallel Task Group API](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F5b63138a650248d4454f306ba72d99dc1572e08a-1600x946.png&w=3840&q=75) ![Image 13: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[The Parallel Task Group API](https://parallel.ai/blog/task-group-api) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/task-group-api) 159 | 160 | [![Image 14: State of the Art Deep Research APIs](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F897a6d4a4d5caf5657b59cdc1dc99de0641c66df-1600x900.png&w=3840&q=75) ![Image 15: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[State of the Art Deep Research APIs](https://parallel.ai/blog/deep-research) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/deep-research) 161 | 162 | [![Image 16: Introducing the Parallel Search API ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Ff61c8461515390c7e546ee3ac1b470a4f79084cc-2560x1704.png&w=3840&q=75) ![Image 17: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Search API ](https://parallel.ai/blog/parallel-search-api) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/parallel-search-api) 163 | 164 | [![Image 18: Introducing the Parallel Chat API - a low latency web research API for web based LLM completions. The Parallel Chat API returns completions in text and structured JSON format, and is OpenAI Chat Completions compatible. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F780ef9d2c09f2e3526a0d11a4bda9eb0a45ef391-2560x1704.jpg&w=3840&q=75) ![Image 19: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Chat API ](https://parallel.ai/blog/chat-api) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/chat-api) 165 | 166 | [![Image 20: Parallel Web Systems introduces Basis with calibrated confidences - a new verification framework for AI web research and search API outputs that sets a new industry standard for transparent and reliable deep research. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F20273a28dbb0a5ee55784f4c9f7050427094f4eb-2672x1512.png&w=3840&q=75) ![Image 21: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Basis with Calibrated Confidences ](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) Tags:Product Release Reading time:4 min](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) 167 | 168 | [![Image 22: The Parallel Task API is a state-of-the-art system for automated web research that delivers the highest accuracy at every price point. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F96491c652ee4988fd9866d1b7619407582879e64-2576x1448.png&w=3840&q=75) ![Image 23: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Task API](https://parallel.ai/blog/parallel-task-api) Tags:Product Release,Benchmarks Reading time:4 min](https://parallel.ai/blog/parallel-task-api) 169 | 170 | ![Company Logo](https://parallel.ai/parallel-logo-540.png) 171 | 172 | ### Company 173 | 174 | * [hello@parallel.ai](mailto:hello@parallel.ai)[hello@parallel.ai](mailto:hello@parallel.ai) 175 | 176 | ### Resources 177 | 178 | * [About](https://parallel.ai/about)[About](https://parallel.ai/about) 179 | * [Docs](https://docs.parallel.ai/)[Docs](https://docs.parallel.ai) 180 | * [Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog) 181 | * [Changelog](https://docs.parallel.ai/resources/changelog)[Changelog](https://docs.parallel.ai/resources/changelog) 182 | * [Careers](https://jobs.ashbyhq.com/parallel)[Careers](https://jobs.ashbyhq.com/parallel) 183 | 184 | ### Info 185 | 186 | * [Terms](https://parallel.ai/terms-of-service)[Terms](https://parallel.ai/terms-of-service) 187 | * [Privacy](https://parallel.ai/privacy-policy)[Privacy](https://parallel.ai/privacy-policy) 188 | * [Trust Center](https://trust.parallel.ai/)[Trust Center](https://trust.parallel.ai/) 189 | 190 | ![SOC 2 Compliant](https://parallel.ai/soc2.svg) 191 | 192 | [LinkedIn](https://www.linkedin.com/company/parallel-web/about/)[LinkedIn](https://www.linkedin.com/company/parallel-web/about/) 193 | 194 | Parallel Web Systems Inc. 2025 195 | -------------------------------------------------------------------------------- /blog/source-policy.md: -------------------------------------------------------------------------------- 1 | Title: Introducing Source Policy 2 | 3 | URL Source: https://parallel.ai/blog/source-policy 4 | 5 | Published Time: 2025-07-08T18:47:48Z 6 | 7 | Markdown Content: 8 | Introducing Source Policy: Precise control over web research sources | Enterprise Deep Research API for ChatGPT & Claude 9 | 10 | =============== 11 | 12 | [Parallel](https://parallel.ai/)[About](https://parallel.ai/about)[About](https://parallel.ai/about)[Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog)[Docs](https://docs.parallel.ai/introduction/quickstart)[Docs](https://docs.parallel.ai/introduction/quickstart) 13 | 14 | [Start Building P](https://platform.parallel.ai/)[Start Building](https://platform.parallel.ai/) 15 | 16 | Menu[Menu] 17 | 18 | Human Machine 19 | 20 | #Introducing Source Policy 21 | ========================== 22 | 23 | Tags:Product Release 24 | 25 | Reading time:1 min 26 | 27 | ![Image 1: Starting today, Source Policy is available for both the Parallel Task API and Search API - giving you granular control over which sources your AI agents access and how results are prioritized.](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9eec7c7314218135df8fd7a83b1c6d3b953eece9-1528x1016.png&w=3840&q=75) 28 | 29 | Starting today, Source Policy is available for both the [Parallel Task API](https://parallel.ai/blog/parallel-task-api)[Parallel Task API]($https://parallel.ai/blog/parallel-task-api) and [Search API](https://parallel.ai/blog/parallel-search-api)[Search API]($https://parallel.ai/blog/parallel-search-api) - giving you granular control over which sources your AI agents access and how results are prioritized. 30 | 31 | ##****What is Source Policy?**** 32 | -------------------------------- 33 | 34 | Source Policy lets you define exactly which domains your research should include and exclude. With two simple parameters, you can shape how Parallel crawls and reasons over the web: 35 | 36 | * -****include_domains****: Restrict queries to specific trusted domains 37 | * -****exclude_domains****: Block unreliable or irrelevant sources 38 | 39 | ###Configure your source policy 40 | 41 | 1 42 | 43 | 2 44 | 45 | 3 46 | 47 | 4 48 | 49 | 5 50 | 51 | 6 52 | 53 | 7 54 | 55 | 8 56 | 57 | 9 58 | 59 | 10 60 | 61 | 11 62 | 63 | 12 64 | 65 | 13 66 | 67 | 14 68 | 69 | ``` 70 | curl -X POST "https://api.parallel.ai/v1/tasks/runs" \ 71 | -H "x-api-key: YOUR_API_KEY" \ 72 | -H "Content-Type: application/json" \ 73 | -H "parallel-beta: source-policy-2025-07-03" \ 74 | -d '{ 75 | "task_spec": { 76 | "output_schema": "How many employees are there in the company" 77 | }, 78 | "input": "Parallel Web Systems", 79 | "processor": "core", 80 | "source_policy": { 81 | "include_domains": ["linkedin.com"] 82 | } 83 | }' 84 | ``` 85 | ```` curl -X POST "https://api.parallel.ai/v1/tasks/runs" \ -H "x-api-key: YOUR_API_KEY" \ -H "Content-Type: application/json" \ -H "parallel-beta: source-policy-2025-07-03" \ -d '{ "task_spec": { "output_schema": "How many employees are there in the company" }, "input": "Parallel Web Systems", "processor": "core", "source_policy": { "include_domains": ["linkedin.com"] } }'```` 86 | 87 | ##****Why Source Policy matters**** 88 | ----------------------------------- 89 | 90 | Production AI applications demand reliable, verifiable sources. Whether you're conducting due diligence that requires SEC filings, sales research that needs company websites, or competitive analysis that excludes certain unreliable publishers - Source Policy ensures your agents access exactly the right information. 91 | 92 | Instead of post-processing results to filter sources, you can now constrain the Parallel web research pipeline from the start. This reduces noise, improves accuracy, and gives you confidence that outputs stem from sources you trust. 93 | 94 | ##****Start building**** 95 | ------------------------ 96 | 97 | Source Policy is available today across all Task API and Search API processors. Get started in our [Developer Platform](https://platform.parallel.ai/)[Developer Platform]($https://platform.parallel.ai) or dive into the [documentation](https://docs.parallel.ai/features/source-policy)[documentation]($https://docs.parallel.ai/features/source-policy). 98 | 99 | ![Image 2: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) 100 | 101 | By Parallel 102 | 103 | July 8, 2025 104 | 105 | ### ##Related Posts 10 106 | 107 | [![Image 3: A linear dithering of a search interface for agents](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fcf3d3b4112c8ab27af1a56442e549a6c9fb2b220-1600x1064.png&w=3840&q=75) ![Image 4: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[A state-of-the-art search API purpose-built for agents](https://parallel.ai/blog/search-api-benchmark) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/search-api-benchmark) 108 | 109 | [![Image 5: Parallel Search MCP Server in Devin](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fdfc00df5061085ec771f28d263b231d4c53747c3-1600x1044.png&w=3840&q=75) ![Image 6: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Parallel Search MCP Server in Devin](https://parallel.ai/blog/parallel-search-mcp-in-devin) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/parallel-search-mcp-in-devin) 110 | 111 | [![Image 7: Introducing Tool Calling via MCP Servers](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fe220ba8a89dfaeb943f943fc92636629a0b9ba14-1600x1600.png&w=3840&q=75) ![Image 8: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Tool Calling via MCP Servers](https://parallel.ai/blog/mcp-tool-calling) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/mcp-tool-calling) 112 | 113 | [![Image 9: Introducing the Parallel Search MCP Server ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F32cf3f1026d452dbaed58e8df4d7b487166d423a-1972x1972.png&w=3840&q=75) ![Image 10: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Search MCP Server ](https://parallel.ai/blog/search-mcp-server) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/search-mcp-server) 114 | 115 | [![Image 11: The Parallel Task Group API](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F5b63138a650248d4454f306ba72d99dc1572e08a-1600x946.png&w=3840&q=75) ![Image 12: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[The Parallel Task Group API](https://parallel.ai/blog/task-group-api) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/task-group-api) 116 | 117 | [![Image 13: State of the Art Deep Research APIs](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F897a6d4a4d5caf5657b59cdc1dc99de0641c66df-1600x900.png&w=3840&q=75) ![Image 14: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[State of the Art Deep Research APIs](https://parallel.ai/blog/deep-research) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/deep-research) 118 | 119 | [![Image 15: Introducing the Parallel Search API ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Ff61c8461515390c7e546ee3ac1b470a4f79084cc-2560x1704.png&w=3840&q=75) ![Image 16: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Search API ](https://parallel.ai/blog/parallel-search-api) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/parallel-search-api) 120 | 121 | [![Image 17: Introducing the Parallel Chat API - a low latency web research API for web based LLM completions. The Parallel Chat API returns completions in text and structured JSON format, and is OpenAI Chat Completions compatible. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F780ef9d2c09f2e3526a0d11a4bda9eb0a45ef391-2560x1704.jpg&w=3840&q=75) ![Image 18: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Chat API ](https://parallel.ai/blog/chat-api) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/chat-api) 122 | 123 | [![Image 19: Parallel Web Systems introduces Basis with calibrated confidences - a new verification framework for AI web research and search API outputs that sets a new industry standard for transparent and reliable deep research. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F20273a28dbb0a5ee55784f4c9f7050427094f4eb-2672x1512.png&w=3840&q=75) ![Image 20: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Basis with Calibrated Confidences ](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) Tags:Product Release Reading time:4 min](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) 124 | 125 | [![Image 21: The Parallel Task API is a state-of-the-art system for automated web research that delivers the highest accuracy at every price point. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F96491c652ee4988fd9866d1b7619407582879e64-2576x1448.png&w=3840&q=75) ![Image 22: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Task API](https://parallel.ai/blog/parallel-task-api) Tags:Product Release,Benchmarks Reading time:4 min](https://parallel.ai/blog/parallel-task-api) 126 | 127 | ![Company Logo](https://parallel.ai/parallel-logo-540.png) 128 | 129 | ### Company 130 | 131 | * [hello@parallel.ai](mailto:hello@parallel.ai)[hello@parallel.ai](mailto:hello@parallel.ai) 132 | 133 | ### Resources 134 | 135 | * [About](https://parallel.ai/about)[About](https://parallel.ai/about) 136 | * [Docs](https://docs.parallel.ai/)[Docs](https://docs.parallel.ai) 137 | * [Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog) 138 | * [Changelog](https://docs.parallel.ai/resources/changelog)[Changelog](https://docs.parallel.ai/resources/changelog) 139 | * [Careers](https://jobs.ashbyhq.com/parallel)[Careers](https://jobs.ashbyhq.com/parallel) 140 | 141 | ### Info 142 | 143 | * [Terms](https://parallel.ai/terms-of-service)[Terms](https://parallel.ai/terms-of-service) 144 | * [Privacy](https://parallel.ai/privacy-policy)[Privacy](https://parallel.ai/privacy-policy) 145 | * [Trust Center](https://trust.parallel.ai/)[Trust Center](https://trust.parallel.ai/) 146 | 147 | ![SOC 2 Compliant](https://parallel.ai/soc2.svg) 148 | 149 | [LinkedIn](https://www.linkedin.com/company/parallel-web/about/)[LinkedIn](https://www.linkedin.com/company/parallel-web/about/) 150 | 151 | Parallel Web Systems Inc. 2025 152 | -------------------------------------------------------------------------------- /blog/task-group-api.md: -------------------------------------------------------------------------------- 1 | Title: The Parallel Task Group API 2 | 3 | URL Source: https://parallel.ai/blog/task-group-api 4 | 5 | Published Time: 2025-07-02T17:56:29Z 6 | 7 | Markdown Content: 8 | Introducing the Parallel Task Group API | Enterprise Deep Research API for ChatGPT & Claude 9 | 10 | =============== 11 | 12 | [Parallel](https://parallel.ai/)[About](https://parallel.ai/about)[About](https://parallel.ai/about)[Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog)[Docs](https://docs.parallel.ai/introduction/quickstart)[Docs](https://docs.parallel.ai/introduction/quickstart) 13 | 14 | [Start Building P](https://platform.parallel.ai/)[Start Building](https://platform.parallel.ai/) 15 | 16 | Menu[Menu] 17 | 18 | Human Machine 19 | 20 | #The Parallel Task Group API 21 | ============================ 22 | 23 | Tags:Product Release 24 | 25 | Reading time:1 min 26 | 27 | ![Image 1: The Parallel Task Group API](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F5b63138a650248d4454f306ba72d99dc1572e08a-1600x946.png&w=3840&q=75) 28 | 29 | [Parallel Tasks](https://parallel.ai/blog/parallel-task-api)[Parallel Tasks]($https://parallel.ai/blog/parallel-task-api) are designed for large-scale workloads. When your pipeline needs to launch hundreds or thousands of independent web research calls, the new ****Task Group API**** wraps those operations into a single batch - giving you one identifier to create, monitor, and collect results from large parallel workloads. 30 | 31 | ###Create a Task Group 32 | 33 | 1 34 | 35 | 2 36 | 37 | 3 38 | 39 | 4 40 | 41 | 5 42 | 43 | 6 44 | 45 | 7 46 | 47 | 8 48 | 49 | 9 50 | 51 | 10 52 | 53 | 11 54 | 55 | 12 56 | 57 | 13 58 | 59 | 14 60 | 61 | 15 62 | 63 | 16 64 | 65 | 17 66 | 67 | 18 68 | 69 | 19 70 | 71 | 20 72 | 73 | 21 74 | 75 | 22 76 | 77 | 23 78 | 79 | 24 80 | 81 | 25 82 | 83 | 26 84 | 85 | 27 86 | 87 | 28 88 | 89 | 29 90 | 91 | 30 92 | 93 | 31 94 | 95 | 32 96 | 97 | 33 98 | 99 | 34 100 | 101 | 35 102 | 103 | 36 104 | 105 | 37 106 | 107 | ``` 108 | import httpx 109 | from parallel.types import TaskRunCreateParams 110 | 111 | client = httpx.Client( 112 | base_url="https://beta.parallel.ai", 113 | headers={"x-api-key": "your-api-key"}, 114 | ) 115 | 116 | # 1. Create a task-group shell 117 | group_resp = client.post("/v1beta/tasks/groups", json={}).json() 118 | taskgroup_id = group_resp["taskgroup_id"] 119 | 120 | # 2. Initiate runs in that group 121 | questions = [ 122 | "What is the capital of France?", 123 | "What is the capital of Germany?", 124 | ] 125 | run_resp = client.post( 126 | f"/v1beta/tasks/groups/{taskgroup_id}/runs", 127 | json={ 128 | "inputs": [ 129 | TaskRunCreateParams(input=q, processor="lite") 130 | for q in questions 131 | ] 132 | }, 133 | ).json() 134 | 135 | print("Initial status:", run_resp["status"]) 136 | 137 | # 3. Stream live events for this run 138 | events_url = f"/v1beta/tasks/groups/{taskgroup_id}/events" 139 | 140 | with client.stream("GET", events_url) as stream: 141 | for chunk in stream.iter_text(): 142 | if chunk.strip(): # ignore keep-alive blanks 143 | print("Event:", chunk) 144 | ``` 145 | ```` import httpxfrom parallel.types import TaskRunCreateParams client = httpx.Client( base_url="https://beta.parallel.ai", headers={"x-api-key": "your-api-key"},) # 1. Create a task-group shellgroup_resp = client.post("/v1beta/tasks/groups", json={}).json()taskgroup_id = group_resp["taskgroup_id"] # 2. Initiate runs in that groupquestions = [ "What is the capital of France?", "What is the capital of Germany?",]run_resp = client.post( f"/v1beta/tasks/groups/{taskgroup_id}/runs", json={ "inputs": [ TaskRunCreateParams(input=q, processor="lite") for q in questions ] },).json() print("Initial status:", run_resp["status"]) # 3. Stream live events for this runevents_url = f"/v1beta/tasks/groups/{taskgroup_id}/events" with client.stream("GET", events_url) as stream: for chunk in stream.iter_text(): if chunk.strip(): # ignore keep-alive blanks print("Event:", chunk) ```` 146 | 147 | ##****Built for Production Scale**** 148 | ------------------------------------ 149 | 150 | Whether you're enriching thousands of CRM records, conducting bulk due diligence, or processing large-scale competitive intelligence workflows - the Task Group API makes running Parallel Tasks in batch seamless. 151 | 152 | ****Unified monitoring**** — Track queued, running, completed, and failed counts through a single endpoint instead of polling hundreds of individual Tasks. 153 | 154 | ****Real-time results streaming**** — Open one connection and receive each Task's structured output the moment it completes, eliminating the need to orchestrate multiple polling loops. 155 | 156 | ****Dynamic expansion**** — Add new Tasks to active groups without restarting batches, supporting workflows that discover additional research targets mid-execution. 157 | 158 | ##****Start Building at Scale**** 159 | --------------------------------- 160 | 161 | The Task Group API is available in public beta today. Get started in our [Developer Platform](https://platform.parallel.ai/)[Developer Platform]($https://platform.parallel.ai/) or dive directly into the [documentation](https://docs.parallel.ai/resources/group-api)[documentation]($https://docs.parallel.ai/resources/group-api). 162 | 163 | ![Image 2: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) 164 | 165 | By Parallel 166 | 167 | July 2, 2025 168 | 169 | ### ##Related Posts 10 170 | 171 | [![Image 3: A linear dithering of a search interface for agents](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fcf3d3b4112c8ab27af1a56442e549a6c9fb2b220-1600x1064.png&w=3840&q=75) ![Image 4: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[A state-of-the-art search API purpose-built for agents](https://parallel.ai/blog/search-api-benchmark) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/search-api-benchmark) 172 | 173 | [![Image 5: Parallel Search MCP Server in Devin](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fdfc00df5061085ec771f28d263b231d4c53747c3-1600x1044.png&w=3840&q=75) ![Image 6: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Parallel Search MCP Server in Devin](https://parallel.ai/blog/parallel-search-mcp-in-devin) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/parallel-search-mcp-in-devin) 174 | 175 | [![Image 7: Introducing Tool Calling via MCP Servers](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Fe220ba8a89dfaeb943f943fc92636629a0b9ba14-1600x1600.png&w=3840&q=75) ![Image 8: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Tool Calling via MCP Servers](https://parallel.ai/blog/mcp-tool-calling) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/mcp-tool-calling) 176 | 177 | [![Image 9: Introducing the Parallel Search MCP Server ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F32cf3f1026d452dbaed58e8df4d7b487166d423a-1972x1972.png&w=3840&q=75) ![Image 10: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Search MCP Server ](https://parallel.ai/blog/search-mcp-server) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/search-mcp-server) 178 | 179 | [![Image 11: Starting today, Source Policy is available for both the Parallel Task API and Search API - giving you granular control over which sources your AI agents access and how results are prioritized.](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9eec7c7314218135df8fd7a83b1c6d3b953eece9-1528x1016.png&w=3840&q=75) ![Image 12: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Source Policy](https://parallel.ai/blog/source-policy) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/source-policy) 180 | 181 | [![Image 13: State of the Art Deep Research APIs](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F897a6d4a4d5caf5657b59cdc1dc99de0641c66df-1600x900.png&w=3840&q=75) ![Image 14: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[State of the Art Deep Research APIs](https://parallel.ai/blog/deep-research) Tags:Benchmarks Reading time:3 min](https://parallel.ai/blog/deep-research) 182 | 183 | [![Image 15: Introducing the Parallel Search API ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2Ff61c8461515390c7e546ee3ac1b470a4f79084cc-2560x1704.png&w=3840&q=75) ![Image 16: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Search API ](https://parallel.ai/blog/parallel-search-api) Tags:Product Release Reading time:2 min](https://parallel.ai/blog/parallel-search-api) 184 | 185 | [![Image 17: Introducing the Parallel Chat API - a low latency web research API for web based LLM completions. The Parallel Chat API returns completions in text and structured JSON format, and is OpenAI Chat Completions compatible. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F780ef9d2c09f2e3526a0d11a4bda9eb0a45ef391-2560x1704.jpg&w=3840&q=75) ![Image 18: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Chat API ](https://parallel.ai/blog/chat-api) Tags:Product Release Reading time:1 min](https://parallel.ai/blog/chat-api) 186 | 187 | [![Image 19: Parallel Web Systems introduces Basis with calibrated confidences - a new verification framework for AI web research and search API outputs that sets a new industry standard for transparent and reliable deep research. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F20273a28dbb0a5ee55784f4c9f7050427094f4eb-2672x1512.png&w=3840&q=75) ![Image 20: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing Basis with Calibrated Confidences ](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) Tags:Product Release Reading time:4 min](https://parallel.ai/blog/introducing-basis-with-calibrated-confidences) 188 | 189 | [![Image 21: The Parallel Task API is a state-of-the-art system for automated web research that delivers the highest accuracy at every price point. ](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F96491c652ee4988fd9866d1b7619407582879e64-2576x1448.png&w=3840&q=75) ![Image 22: Parallel avatar](https://parallel.ai/_next/image?url=https%3A%2F%2Fcdn.sanity.io%2Fimages%2F5hzduz3y%2Fproduction%2F9a2c0f2e9634a95512da83f8354aef9d5bf400aa-128x128.png&w=64&q=75) ### -[Introducing the Parallel Task API](https://parallel.ai/blog/parallel-task-api) Tags:Product Release,Benchmarks Reading time:4 min](https://parallel.ai/blog/parallel-task-api) 190 | 191 | ![Company Logo](https://parallel.ai/parallel-logo-540.png) 192 | 193 | ### Company 194 | 195 | * [hello@parallel.ai](mailto:hello@parallel.ai)[hello@parallel.ai](mailto:hello@parallel.ai) 196 | 197 | ### Resources 198 | 199 | * [About](https://parallel.ai/about)[About](https://parallel.ai/about) 200 | * [Docs](https://docs.parallel.ai/)[Docs](https://docs.parallel.ai) 201 | * [Blog](https://parallel.ai/blog)[Blog](https://parallel.ai/blog) 202 | * [Changelog](https://docs.parallel.ai/resources/changelog)[Changelog](https://docs.parallel.ai/resources/changelog) 203 | * [Careers](https://jobs.ashbyhq.com/parallel)[Careers](https://jobs.ashbyhq.com/parallel) 204 | 205 | ### Info 206 | 207 | * [Terms](https://parallel.ai/terms-of-service)[Terms](https://parallel.ai/terms-of-service) 208 | * [Privacy](https://parallel.ai/privacy-policy)[Privacy](https://parallel.ai/privacy-policy) 209 | * [Trust Center](https://trust.parallel.ai/)[Trust Center](https://trust.parallel.ai/) 210 | 211 | ![SOC 2 Compliant](https://parallel.ai/soc2.svg) 212 | 213 | [LinkedIn](https://www.linkedin.com/company/parallel-web/about/)[LinkedIn](https://www.linkedin.com/company/parallel-web/about/) 214 | 215 | Parallel Web Systems Inc. 2025 216 | -------------------------------------------------------------------------------- /fetch.js: -------------------------------------------------------------------------------- 1 | import { readFile, writeFile, mkdir } from "fs/promises"; 2 | import { existsSync } from "fs"; 3 | import { dirname } from "path"; 4 | 5 | async function fetchAndSaveUrls() { 6 | try { 7 | // Read the README.md file 8 | const readmeContent = await readFile("urls.md", "utf-8"); 9 | 10 | // Split into lines and filter out empty lines 11 | const urls = readmeContent 12 | .split("\n") 13 | .map((line) => line.trim()) 14 | .filter((line) => line && line.startsWith("http")); 15 | 16 | console.log(`Found ${urls.length} URLs to fetch`); 17 | 18 | // Process each URL 19 | for (const url of urls) { 20 | try { 21 | console.log(`Fetching: ${url}`); 22 | 23 | // Fetch the URL 24 | const response = await fetch("https://r.jina.ai/" + url); 25 | 26 | if (!response.ok) { 27 | console.error( 28 | `Failed to fetch ${url}: ${response.status} ${response.statusText}` 29 | ); 30 | continue; 31 | } 32 | 33 | // Get the content 34 | const content = await response.text(); 35 | 36 | // Extract the last segment of the pathname 37 | const urlObj = new URL(url); 38 | const lastSegment = urlObj.pathname.slice(1) || "index"; 39 | const filename = `${lastSegment}.md`; 40 | 41 | // Ensure the directory exists 42 | const dir = dirname(filename); 43 | if (dir !== "." && !existsSync(dir)) { 44 | await mkdir(dir, { recursive: true }); 45 | console.log(`Created directory: ${dir}`); 46 | } 47 | 48 | // Save to file 49 | await writeFile(filename, content, "utf-8"); 50 | console.log(`Saved: ${filename}`); 51 | } catch (error) { 52 | console.error(`Error processing ${url}:`, error.message); 53 | } 54 | } 55 | 56 | console.log("Finished processing all URLs"); 57 | } catch (error) { 58 | console.error("Error reading README.md:", error.message); 59 | } 60 | } 61 | 62 | // Run the script 63 | fetchAndSaveUrls(); 64 | -------------------------------------------------------------------------------- /platform.parallel.ai/pricing.md: -------------------------------------------------------------------------------- 1 | # Parallel Pricing Overview 2 | 3 | The Parallel API has usage-based pricing and is priced per 1000 requests. 4 | 5 | ## Available Task API Processors 6 | 7 | Each of our processors are priced based on the complexity of the research Task. A single request corresponds to a single set of inputs converted to outputs. For specialized processor access, contact us. 8 | 9 | | Tier | Price | Description | 10 | | --------- | ------------------------- | ------------------------------ | 11 | | **Lite** | $5.00 per 1000 requests | Basic information retrieval | 12 | | **Base** | $10.00 per 1000 requests | Simple web research | 13 | | **Core** | $25.00 per 1000 requests | Complex web research | 14 | | **Pro** | $100.00 per 1000 requests | Exploratory web research | 15 | | **Ultra** | $300.00 per 1000 requests | Advanced multi-source research | 16 | 17 | ## Available Chat API Processors 18 | 19 | | Tier | Price | Description | 20 | | --------- | ----------------------- | --------------------------------------- | 21 | | **Speed** | $5.00 per 1000 requests | Low-latency, web-based chat completions | 22 | 23 | ## Available Search API Processors 24 | 25 | | Tier | Price | Description | 26 | | -------- | ----------------------- | ------------------------------------------------- | 27 | | **Base** | $4.00 per 1000 requests | Low latency structured web extraction | 28 | | **Pro** | $9.00 per 1000 requests | Structured web extraction optimized for freshness | 29 | -------------------------------------------------------------------------------- /privacy-policy.md: -------------------------------------------------------------------------------- 1 | Title: Privacy Policy | Enterprise Deep Research API for ChatGPT & Claude 2 | 3 | URL Source: https://parallel.ai/privacy-policy 4 | 5 | Markdown Content: 6 | Effective date:April 7, 2025 7 | 8 | Thank you for visiting www.parallel.ai (the “Sites”) owned and operated by Parallel. At Parallel, we take your privacy seriously. Please read this Privacy Policy to learn how we treat your personal data. 9 | 10 | By using or accessing our Sites in any manner, you acknowledge that you accept the practices and policies outlined below, and you hereby consent that we will collect, use and disclose your information as described in this Privacy Policy. 11 | 12 | Remember that your use of Parallel Services is at all times subject to our Terms of Use, which incorporates this Privacy Policy. Any terms we use in this Policy without defining them have the definitions given to them in the Terms of Use. 13 | 14 | You may print a copy of this Privacy Policy by clicking here. 15 | 16 | As we continually work to improve our Services, we may need to change this Privacy Policy from time to time. We will alert you of material changes by placing a notice on the Parallel website, by sending you an email and/or by some other means. Please note that if you’ve opted not to receive legal notice emails from us (or you haven’t provided us with your email address), those legal notices will still govern your use of the Services, and you are still responsible for reading and understanding them. If you use the Services after any changes to the Privacy Policy have been posted, that means you agree to all of the changes. 17 | 18 | This Privacy Policy covers how we treat Personal Data that we gather when you access or use our Services. “Personal Data” means any information that identifies or relates to a particular individual and also includes information referred to as “personally identifiable information” or “personal information” or “sensitive personal information” under applicable data privacy laws, rules or regulations. This Privacy Policy does not cover the practices of companies we don’t own or control or people we don’t manage. 19 | 20 | ### ###Categories of Personal Data We Collect 21 | 22 | This list details the categories of Personal Data that we collect and have collected over the past 12 months: 23 | 24 | #### ####1. Profile or Contact Data 25 | 26 | Examples: 27 | 28 | * -First and last name, email, phone number 29 | 30 | Business or Commercial Purpose(s) for Collection: 31 | 32 | * -Providing, Customizing and Improving the Services 33 | * -Corresponding with You 34 | 35 | Third Parties With Whom We Disclose this Personal Data: 36 | 37 | * -Service Providers 38 | 39 | #### ####2. Device/IP Data 40 | 41 | Examples: 42 | 43 | * -IP address, type of device/operating system/browser used to access the Services 44 | 45 | Business or Commercial Purpose(s) for Collection: 46 | 47 | * -Providing, Customizing and Improving the Services 48 | * -Corresponding with You 49 | 50 | Third Parties With Whom We Disclose this Personal Data: 51 | 52 | * -Service Providers 53 | 54 | #### ####3. Web Analytics 55 | 56 | Examples: 57 | 58 | * -Web page interactions, referring webpage/source, statistics associated with the interaction between device or browser and the Services 59 | 60 | Business or Commercial Purpose(s) for Collection: 61 | 62 | * -Providing, Customizing and Improving the Services 63 | * -Corresponding with You 64 | 65 | Third Parties With Whom We Disclose this Personal Data: 66 | 67 | * -Service Providers 68 | 69 | #### ####4. Other Identifying Information 70 | 71 | Examples: 72 | 73 | * -Emails, letters, texts, or other communications you send us 74 | 75 | Business or Commercial Purpose(s) for Collection: 76 | 77 | * -Providing, Customizing and Improving the Services 78 | * -Corresponding with You 79 | 80 | Third Parties With Whom We Disclose this Personal Data: 81 | 82 | * -Service Providers 83 | 84 | #### ####Our commercial or business purposes for collecting personal data 85 | 86 | 1. Providing, Customizing and Improving the Services 87 | 88 | * -Creating and managing your account or other user profiles. 89 | * -Providing you with the products, services or information you request. 90 | * -Meeting or fulfilling the reason you provided the information to us. 91 | * -Providing support and assistance for the Services. 92 | * -Improving the Services, including testing, research, internal analytics and product development. 93 | * -Personalizing the Services, website content and communications based on your preferences. 94 | * -Doing fraud protection, security and debugging. 95 | 96 | #### ####2. Corresponding with You 97 | 98 | * -Responding to correspondence that we receive from you, contacting you when necessary or requested, and sending you information about the Company or the Services. 99 | * -Sending emails and other communications according to your preferences 100 | 101 | ### ###Our Permitted Purposes for Processing Personal Data 102 | 103 | In addition, each of the above referenced categories of Personal Data may be collected, used, and disclosed with the government, including law enforcement, or other parties to meet certain legal requirements and enforcing legal terms including: fulfilling our legal obligations under applicable law, regulation, court order or other legal process, such as preventing, detecting and investigating security incidents and potentially illegal or prohibited activities; protecting the rights, property or safety of you, Parallel or another party; enforcing any agreements with you; responding to claims that any posting or other content violates third-party rights; and resolving disputes. 104 | 105 | We will not collect additional categories of Personal Data or use the Personal Data we collected for materially different, unrelated or incompatible purposes without providing you notice or obtaining your consent. 106 | 107 | ### ###Categories of Sources of Personal Data 108 | 109 | We collect Personal Data about you from the following categories of sources: 110 | 111 | #### ####You 112 | 113 | When you provide such information directly to us. 114 | 115 | * -When you use our interactive tools and Services. 116 | * -When you voluntarily provide information in free-form text boxes through the Services or through responses to surveys or questionnaires. 117 | * -When you send us an email or otherwise contact us. 118 | 119 | When you use the Services and such information is collected automatically. 120 | 121 | * -Through Cookies (defined in the “Tracking Tools, Advertising and Opt-Out” section below). 122 | * -If you use a location-enabled browser, we may receive information about your location. 123 | 124 | #### ####Third Parties 125 | 126 | Vendors 127 | 128 | * -We may use analytics providers to analyze how you interact and engage with the Services, or third parties may help us provide you with customer support. 129 | 130 | We disclose your Personal Data to the categories of service providers and other parties listed in this section. Depending on state laws that may be applicable to you, some of these disclosures may constitute a “sale” of your Personal Data. For more information, please refer to the state-specific sections below. 131 | 132 | ### ###Service Providers 133 | 134 | These parties help us provide the Services or perform business functions on our behalf. They include: 135 | 136 | * -Hosting, technology and communication providers. 137 | * -Security and fraud prevention consultants. 138 | * -Support and customer service vendors. 139 | * -Analytics providers regarding web traffic or usage of the Services. 140 | 141 | ### ###Legal Obligations 142 | 143 | We may disclose any Personal Data that we collect with third parties in conjunction with any of the activities set forth under “Other Permitted Purposes for Processing Personal Data” section above. 144 | 145 | ### ###Business Transfers 146 | 147 | All of your Personal Data that we collect may be transferred to a third party if we undergo a merger, acquisition, bankruptcy or other transaction in which that third party assumes control of our business (in whole or in part). 148 | 149 | ### ###Data that is Not Personal Data 150 | 151 | We may create aggregated, de-identified or anonymized data from the Personal Data we collect, including by removing information that makes the data personally identifiable to a particular user. We may use such aggregated, de-identified or anonymized data and disclose it with third parties for our lawful business purposes, including to analyze, build and improve the Services and promote our business, provided that we will not disclose such data in a manner that could identify you. 152 | 153 | The Services use cookies and similar technologies such as pixel tags, web beacons, clear GIFs and JavaScript (collectively, “Cookies”) to enable our servers to recognize your web browser, tell us how and when you visit and use our Services, analyze trends, learn about our user base and operate and improve our Services. Cookies are small pieces of data– usually text files – placed on your computer, tablet, phone or similar device when you use that device to access our Services. We may also supplement the information we collect from you with information received from third parties, including third parties that have placed their own Cookies on your device(s). 154 | 155 | Please note that because of our use of Cookies, the Services do not support “Do Not Track” requests sent from a browser at this time. 156 | 157 | We use the following types of Cookies: 158 | 159 | * -Essential Cookies are required for providing you with features or services that you have requested. For example, certain Cookies enable you to log into secure areas of our Services. Disabling these Cookies may make certain features and services unavailable. 160 | * -Functional Cookies are used to record your choices and settings regarding our Services, maintain your preferences over time and recognize you when you return to our Services. These Cookies help us to personalize our content for you, greet you by name and remember your preferences (for example, your choice of language or region). 161 | * -Performance/Analytical Cookies allow us to understand how visitors use our Services. They do this by collecting information about the number of visitors to the Services, what pages visitors view on our Services and how long visitors are viewing pages on the Services. Performance/Analytical Cookies also help us measure the performance of our advertising campaigns in order to help us improve our campaigns and the Services’ content for those who engage with our advertising. For example, Google LLC (“Google”) uses cookies in connection with its Google Analytics services. Google’s ability to use and disclose information collected by Google Analytics about your visits to the Services is subject to the Google Analytics Terms of Use and the Google Privacy Policy. You have the option to opt-out of Google’s use of Cookies by visiting the Google advertising opt-out page at www.google.com/privacy_ads.html or the Google Analytics Opt-out Browser Add-on at https://tools.google.com/dlpage/gaoptout/. 162 | 163 | You can decide whether or not to accept Cookies through your internet browser’s settings. Most browsers have an option for turning off the Cookie feature, which will prevent your browser from accepting new Cookies, as well as (depending on the sophistication of your browser software) allow you to decide on acceptance of each new Cookie in a variety of ways. You can also delete all Cookies that are already on your device. If you do this, however, you may have to manually adjust some preferences every time you visit our website and some of the Services and functionalities may not work. 164 | 165 | To explore what Cookie settings are available to you or to modify your preferences with respect to Cookies, you can access your Cookie management settings by clicking [LINK]. To find out more information about Cookies generally, including information about how to manage and delete Cookies, please visit http://www.allaboutcookies.org/. 166 | 167 | We seek to protect your Personal Data from unauthorized access, use and disclosure using appropriate physical, technical, organizational and administrative security measures based on the type of Personal Data and how we are processing that data. You should also help protect your data by appropriately selecting and protecting your password and/or other sign-on mechanism; limiting access to your computer or device and browser; and signing off after you have finished accessing your account. Although we work to protect the security of your account and other data that we hold in our records, please be aware that no method of transmitting data over the internet or storing data is completely secure. 168 | 169 | We retain Personal Data about you for as long as necessary to provide you with our Services or to perform our business or commercial purposes for collecting your Personal Data. When establishing a retention period for specific categories of data, we consider who we collected the data from, our need for the Personal Data, why we collected the Personal Data, and the sensitivity of the Personal Data. In some cases we retain Personal Data for longer, if doing so is necessary to comply with our legal obligations, resolve disputes or collect fees owed, or is otherwise permitted or required by applicable law, rule or regulation. We may further retain information in an anonymous or aggregated form where that information would not identify you personally. 170 | 171 | For example: 172 | 173 | * -We retain your profile information and credentials for as long as you have an account with us. 174 | * -We retain user device/IP data for as long as we need it to ensure that our systems are working appropriately, effectively and efficiently. 175 | 176 | As noted in the Terms of Use, we do not knowingly collect or solicit Personal Data from children under 16 years of age; if you are a child under the age of 16, please do not attempt to register for or otherwise use the Services or send us any Personal Data. If we learn we have collected Personal Data from a child under 16 years of age, we will delete that information as quickly as possible. If you believe that a child under 16 years of age may have provided Personal Data to us, please contact us at support@parallel.ai. 177 | 178 | ### ###California Resident Rights 179 | 180 | Under California Civil Code Sections 1798.83-1798.84, California residents are entitled to contact us to prevent disclosure of Personal Data to third parties for such third parties’ direct marketing purposes; in order to submit such a request, please contact us at support@parallel.ai. 181 | 182 | Your browser may offer you a “Do Not Track” option, which allows you to signal to operators of websites and web applications and services that you do not wish such operators to track certain of your online activities over time and across different websites. Our Services do not support Do Not Track requests at this time. To find out more about “Do Not Track,” you can visit [www.allaboutdnt.com](http://www.allaboutdnt.com/)[www.allaboutdnt.com]($http://www.allaboutdnt.com/). 183 | 184 | If you are a resident of Nevada, you have the right to opt-out of the sale of certain Personal Data to third parties. Please note that we do not currently sell your Personal Data as sales are defined in Nevada Revised Statutes Chapter 603A. 185 | 186 | If you have any questions or comments about this Privacy Policy, the ways in which we collect and use your Personal Data or your choices and rights regarding such collection and use, please do not hesitate to contact us at: 187 | 188 | [support@parallel.ai](mailto:support@parallel.ai)[support@parallel.ai]($mailto:support@parallel.ai) 189 | 190 | 2261 Market Street, #5578, San Francisco, CA 941 191 | -------------------------------------------------------------------------------- /terms-of-service.md: -------------------------------------------------------------------------------- 1 | Title: Terms | Enterprise Deep Research API for ChatGPT & Claude 2 | 3 | URL Source: https://parallel.ai/terms-of-service 4 | 5 | Markdown Content: 6 | Effective date:April 7, 2025 7 | 8 | Welcome to Parallel. Please read on to learn the rules and restrictions that govern your use of our websites. If you have any questions, comments, or concerns regarding these terms or our website, please contact us at [support@parallel.ai](mailto:support@parallel.ai)[support@parallel.ai]($mailto:support@parallel.ai). 9 | 10 | Address: 2261 Market Street, #5578, San Francisco, CA 94114 11 | 12 | These Terms of Use (the “Terms”) are a binding contract between you and PARALLEL WEB SYSTEMS INC. (“Parallel,” “we” and “us”). Your use of our websites in any way means that you agree to all of these Terms, and these Terms will remain in effect while you use our websites. These Terms include the provisions in this document as well as those in the Privacy Policy. 13 | 14 | Please read these Terms carefully. They cover important information about our website. PLEASE NOTE THAT YOUR USE OF AND ACCESS TO OUR WEBSITES ARE SUBJECT TO THE FOLLOWING TERMS; IF YOU DO NOT AGREE TO ALL OF THE FOLLOWING, YOU MAY NOT USE OR ACCESS OUR WEBSITES IN ANY MANNER. 15 | 16 | We are constantly trying to improve our Services, so these Terms may need to change along with our Services. We reserve the right to change the Terms at any time, but if we do, we will place a notice on our site located at parallel.ai, send you an email, and/or notify you by some other means. 17 | 18 | If you don’t agree with the new Terms, you are free to reject them; unfortunately, that means you will no longer be able to use the Services. If you use the Services in any way after a change to the Terms is effective, that means you agree to all of the changes. 19 | 20 | Except for changes by us as described here, no other amendment or modification of these Terms will be effective unless in writing and signed by both you and us. 21 | 22 | Parallel takes the privacy of its users very seriously. For the current Parallel Privacy Policy, please [click here](https://www.parallel.ai/privacy-policy)[click here]($https://www.parallel.ai/privacy-policy). 23 | 24 | ### ###Children’s Online Privacy Protection Act 25 | 26 | The Children’s Online Privacy Protection Act (“COPPA”) requires that online service providers obtain parental consent before they knowingly collect personally identifiable information online from children who are under 16 years of age. We do not knowingly collect or solicit personally identifiable information from children under 16 years of age; if you are a child under 16 years of age, please do not attempt to register for or otherwise use our websites or send us any personal information. If we learn we have collected personal information from a child under 16 years of age, we will delete that information as quickly as possible. If you believe that a child under 16 years of age may have provided us personal information, please contact us at support@parallel.ai. 27 | 28 | You represent and warrant that you are an individual of legal age to form a binding contract (or if not, you’ve received your parent’s or guardian’s permission to use our websites and have gotten your parent or guardian to agree to these Terms on your behalf). If you’re agreeing to these Terms on behalf of an organization or entity, you represent and warrant that you are authorized to agree to these Terms on that organization’s or entity’s behalf and bind them to these Terms (in which case, the references to “you” and “your” in these Terms, except for in this sentence, refer to that organization or entity). 29 | 30 | You will only use our websites in a manner that complies with all laws that apply to you. If your use of our websites is prohibited by applicable laws, then you aren’t authorized to use our website. We can’t and won’t be responsible for your using our websites in a way that breaks the law. 31 | 32 | As part of our websites, you may receive communications from us, including messages that Parallel sends you (for example, via email). 33 | 34 | You represent, warrant, and agree that you will not provide or contribute anything, including any Content (as that term is defined below), to our websites, or otherwise use or interact with our websites, in a manner that: 35 | 36 | * -infringes or violates the intellectual property rights or any other rights of anyone else (including Parallel); 37 | * -violates any law or regulation, including, without limitation, any applicable export control laws, privacy laws or any other purpose not reasonably intended by Parallel; 38 | * -is dangerous, harmful, fraudulent, deceptive, threatening, harassing, defamatory, obscene, or otherwise objectionable; 39 | * -attempts, in any manner, to obtain the password, account, or other security information from any other user; 40 | * -violates the security of any computer network, or cracks any passwords or security encryption codes; 41 | * -runs Maillist, Listserv, any form of auto-responder or “spam” on our websites, or that otherwise interfere with the proper working of our websites (including by placing an unreasonable load on the websites’ infrastructure); 42 | * -“crawls,” “scrapes,” or “spiders” any page, data, or portion of or relating to our websites or Content (through use of manual or automated means); 43 | * -copies or stores any significant portion of the Content; or 44 | * -decompiles, reverse engineers, or otherwise attempts to obtain the source code or underlying ideas or information of or relating to our websites. 45 | 46 | A violation of any of the foregoing is grounds for termination of your right to use or access our websites. 47 | 48 | The materials displayed or performed or available on or through our website, including, but not limited to, text, graphics, data, articles, photos, images, illustrations and so forth (all of the foregoing, the “Content”) are protected by copyright and/or other intellectual property laws. You promise to abide by all copyright notices, trademark rules, information, and restrictions contained in any Content you access through our website, and you won’t use, copy, reproduce, modify, translate, publish, broadcast, transmit, distribute, perform, upload, display, license, sell, commercialize or otherwise exploit for any purpose any Content not owned by you, (i) without the prior consent of the owner of that Content or (ii) in a way that violates someone else’s (including Parallel's) rights. 49 | 50 | Subject to these Terms, we grant each user of our websites a worldwide, non-exclusive, non-sublicensable and non-transferable license to use (i.e., to download and display locally) Content solely for purposes of using our website. Use, reproduction, modification, distribution or storage of any Content for any purpose other than using our websites is expressly prohibited without prior written permission from us. You understand that Parallel owns our website. You won’t modify, publish, transmit, participate in the transfer or sale of, reproduce (except as expressly provided in this Section), create derivative works based on, or otherwise exploit any of our website. Our websites may allow you to copy or download certain Content, but please remember that even where these functionalities exist, all the restrictions in this section still apply. 51 | 52 | Warranty Disclaimer. Neither Parallel nor its licensors, suppliers, partners, parent, subsidiaries or affiliated entities, and each of their respective officers, directors, members, employees, consultants, contract employees, representatives and agents, or any of their respective successors and assigns (Parallel and all such parties together, the “Parallel Parties”) make any representations or warranties concerning our websites or any subject matter hereof , including without limitation regarding any content, services or products contained in or accessed through our website. . PARALLEL FOR ITSELF AND THE PARALLEL PARTIES DISCLAIMS ALL WARRANTIES, OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING, WITHOUT LIMITATION, IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, NON-INFRINGEMENT,. SOME STATES DO NOT ALLOW LIMITATIONS ON HOW LONG AN IMPLIED WARRANTY LASTS, SO THE ABOVE LIMITATIONS MAY NOT APPLY TO YOU. 53 | 54 | Limitation of Liability. TO THE FULLEST EXTENT ALLOWED BY APPLICABLE LAW, UNDER NO CIRCUMSTANCES AND UNDER NO LEGAL THEORY (INCLUDING, WITHOUT LIMITATION, TORT, CONTRACT, STRICT LIABILITY, OR OTHERWISE) SHALL ANY OF THE PARALLEL PARTIES BE LIABLE TO YOU OR TO ANY OTHER PERSON FOR (A) ANY INDIRECT, SPECIAL, INCIDENTAL, PUNITIVE OR CONSEQUENTIAL DAMAGES OF ANY KIND, INCLUDING DAMAGES FOR LOST PROFITS, BUSINESS INTERRUPTION, LOSS OF DATA, LOSS OF GOODWILL, WORK STOPPAGE, ACCURACY OF RESULTS, OR COMPUTER FAILURE OR MALFUNCTION, (B) ANY SUBSTITUTE GOODS, SERVICES OR TECHNOLOGY, (C) ANY AMOUNT, IN THE AGGREGATE, IN EXCESS OF ONE-HUNDRED ($100) DOLLARS, OR (D) ANY MATTER BEYOND OUR REASONABLE CONTROL. SOME STATES DO NOT ALLOW THE EXCLUSION OR LIMITATION OF INCIDENTAL OR CONSEQUENTIAL OR CERTAIN OTHER DAMAGES, SO THE ABOVE LIMITATION AND EXCLUSIONS MAY NOT APPLY TO YOU. 55 | 56 | Assignment. You may not assign, delegate or transfer these Terms or your rights or obligations hereunder, in any way (by operation of law or otherwise) without Parallel's prior written consent. We may transfer, assign, or delegate these Terms and our rights and obligations without consent. 57 | 58 | Choice of Law. These Terms are governed by and will be construed under applicable federal law, and the laws of the State of California, without regard to the conflicts of laws provisions thereof. 59 | 60 | Miscellaneous. The failure of either you or us to exercise, in any way, any right herein shall not be deemed a waiver of any further rights hereunder. If any provision of these Terms are found to be unenforceable or invalid, that provision will be limited or eliminated, to the minimum extent necessary, so that these Terms shall otherwise remain in full force and effect and enforceable. You and Parallel agree that these Terms are the complete and exclusive statement of the mutual understanding between you and Parallel, and that these Terms supersede and cancel all previous written and oral agreements, communications and other understandings relating to the subject matter of these Terms. You hereby acknowledge and agree that you are not an employee, agent, partner, or joint venture of Parallel, and you do not have any authority of any kind to bind Parallel in any respect whatsoever. 61 | 62 | You and Parallel agree there are no third-party beneficiaries intended under these Terms. 63 | -------------------------------------------------------------------------------- /urls.md: -------------------------------------------------------------------------------- 1 | https://parallel.ai/blog/parallel-task-api 2 | https://parallel.ai/blog/introducing-basis-with-calibrated-confidences 3 | https://parallel.ai/blog/chat-api 4 | https://parallel.ai/blog/parallel-search-api 5 | https://parallel.ai/blog/deep-research 6 | https://parallel.ai/blog/task-group-api 7 | https://parallel.ai/blog/source-policy 8 | https://parallel.ai/blog/search-mcp-server 9 | https://parallel.ai/blog/parallel-search-mcp-in-devin 10 | https://parallel.ai/blog/search-api-benchmark 11 | https://parallel.ai/privacy-policy 12 | https://parallel.ai/terms-of-service 13 | https://parallel.ai/about 14 | --------------------------------------------------------------------------------