With Parallel retrieval
How Uber and Lyft won their first million users
Uber
Lyft
Watching 1,200 Accounts
Company
Product releases
”Authoritative legal data across dozens of countries lives on sites no search engine has ever indexed. Parallel solves that at a scale we couldn't build ourselves.”
Where agents find answers
Web infrastructure for AI to search, extract, monitor, and reason over the world's information
TAB
TAB
Fetch
TAB
fromTAB
Join the teams building the AI frontier
Trusted by category-defining startups & Fortune 500 enterprises
- Harvey
- Formation Bio
- Attio
- Starbridge
- Granola
- Monaco
- Manus
- Hex
- Modal
- Dropbox
- Owner
- Greptile
- Rogo
- Profound
- Opendoor
Join the teams building the AI frontier
Trusted by category-defining startups & Fortune 500 enterprises
Put the web to work across every industry
Monitor portfolios, signals, and pricing
Watch the things your business runs on: portfolio companies, target accounts, competitor pricing, regulations. Get a structured alert the moment something material changes, so you only hear what matters to you.
Example query: Find the latest funding rounds for AI infrastructure startups
Answer with fresh facts, inside your product
Give your assistant, copilot, or coding agent current, cited answers fast enough for a chat thread: live docs, real-world context, breaking news. Your users get the answer that's relevant today, not at training time.
Example query: What changed in the EU AI Act this month?
Automate research on any account, market, or thesis at scale
Hand over an objective and get back hundreds of finished, cited reports in minutes: an account brief before the call, an investment memo, a competitive landscape. Analyst-grade work, embedded in your workflow.
Example query: Build a cited brief on the industrial robotics market
Enrich companies, people, and accounts
Drop in a list or describe who you're looking for. Get back live, structured profiles: target accounts, investment candidates, vendors, or the long tail. Every field cited and confidence-scored, ready for your CRM or pipeline.
Example query: Enrich these 50 accounts with headcount and funding
Put the web to work across every industry
Watch the things your business runs on: portfolio companies, target accounts, competitor pricing, regulations. Get a structured alert the moment something material changes, so you only hear what matters to you.
Give your assistant, copilot, or coding agent current, cited answers fast enough for a chat thread: live docs, real-world context, breaking news. Your users get the answer that's relevant today, not at training time.
Hand over an objective and get back hundreds of finished, cited reports in minutes: an account brief before the call, an investment memo, a competitive landscape. Analyst-grade work, embedded in your workflow.
Drop in a list or describe who you're looking for. Get back live, structured profiles: target accounts, investment candidates, vendors, or the long tail. Every field cited and confidence-scored, ready for your CRM or pipeline.
Better answers start with better web search
Parallel gives agents real-time access to fresh and relevant context from the web. Search billions of pages, extract key information, monitor changes, and do complex research with cited outputs you can trust in production.
What do I need to know for my meeting tomorrow?
With Parallel retrieval
The company closed a $40M Series B last week and has been hiring aggressively in enterprise sales since, a signal they're moving upmarket. The lead contact joined as VP of Product in January, previously led product at a direct competitor, and published a piece last week flagging retrieval quality as their biggest infrastructure gap.
High confidence
LLM only
The company is an enterprise SaaS business focused on workflow automation, backed by Sequoia and Index Ventures. The lead contact is a senior product leader who previously worked at Salesforce. They've been growing their enterprise segment and expanding internationally.
Better answers start with better web search
Parallel gives agents real-time access to fresh and relevant context from the web. Search billions of pages, extract key information, monitor changes, and do complex research with cited outputs you can trust in production.
What do I need to know for my meeting tomorrow?
LLM only
Your agent
The company is an enterprise SaaS business focused on workflow automation, backed by Sequoia and Index Ventures. The lead contact is a senior product leader who previously worked at Salesforce. They've been growing their enterprise segment and expanding internationally.
With Parallel retrieval
Your agent
The company closed a $40M Series B last week and has been hiring aggressively in enterprise sales since, a signal they're moving upmarket. The lead contact joined as VP of Product in January, previously led product at a direct competitor, and published a piece last week flagging retrieval quality as their biggest infrastructure gap.
High confidence
Designed to compose
With Parallel, agents can find quick answers, watch the web for changes, enrich databases, and do deep research across the deepest corners of the web, all through a single API.
- EXTRACT: Find the latest news on my competitors
- MONITOR: Alert me when any of them raises a new round
- DEEP RESEARCH: Build a cited funding report
EXTRACT
Pull funding details from these 40 press releases
MONITOR
Alert me when any of them raises a new round
DEEP RESEARCH
Build a cited brief on the most active investors
Designed to compose
With Parallel, agents can find quick answers, watch the web for changes, enrich databases, and do deep research across the deepest corners of the web, all through a single API.
Production-grade trust
Parallel's Web Agents use Basis to attach provenance and calibrated confidence scoring to every output, for trusted use and auditability in production. Every score can be fed back into retrieval to continuously improve results over time.
Production-grade trust
Parallel's Web Agents use Basis to attach provenance and calibrated confidence scoring to every output, for trusted use and auditability in production. Every score can be fed back into retrieval to continuously improve results over time.
More signal per token
Reduce your inference spend while improving quality with more context-efficient retrieval. Every query is served by Parallel’s web index, purpose-built for agents.
More signal per token
Reduce your inference spend while improving quality with more context-efficient retrieval. Every query is served by Parallel’s web index, purpose-built for agents.
Products
Everything your agent needs to work with the web
From a single search to always-on monitoring, Parallel gives agents every way to find, research, and act on web data. All on one platform, powered by the same proprietary index.
Task
Deep research and analysis at enterprise scale. Give it an objective, get back structured, cited answers
Monitor
Track changes and act on what matters, with structured alerts via webhook.
Responses
Cited, synthesized answers from the live web in seconds, built for synchronous interactions.
Search
The highest-accuracy web search for your AI. Fresh, fast, and affordable grounding in one call.
Extract
Get full or excerpted contents from the public web, including PDFs, JS-heavy pages, and government sites
FindAll
Build structured datasets with real-time enrichment. Describe what you want, get a live list.
Retrieval, built from the ground up
Parallel makes the web programmable, bringing the world’s information to agents, grounding them in better facts, helping them do complex knowledge work, and even keeping them proactively informed. Everything runs on Parallel’s proprietary web search stack, built for the scale AI demands.
Web Agents
Specialized agents for web research, built to outperform generalist models on cost, quality, and latency. They plan each question, search, read across dense sources, and return cited answers, all running on custom models, a purpose-built harness, and the Parallel Web Index.
Web Tools
Raw web search primitives designed from the ground up for use by AI agents, including our own. Search finds and ranks the most relevant information sources to ground answers in real-time data; Extract fetches content directly from the page for easy LLM synthesis.
Web Index
Proprietary, state-of-the-art index of the web, continuously updated across hundreds of billions of pages and ranked for what’s authoritative and fresh. It includes licensed content other indexes don’t have: paywalled publishers, premium newsletters, and proprietary sources through Index partnerships. Every search call, every agent runs on it.
Our products
Retrieval, built from the ground up
Parallel makes the web programmable, bringing the world's information to agents, grounding them in better facts, helping them do complex knowledge work, and even keeping them pro-actively informed. Everything runs on Parallel's proprietary web search stack, built for the scale AI demands.
Deep research and analysis at enterprise scale. Give it an objective, get back structured, cited answers
Track changes and act on what matters, with structured alerts via webhook.
Build structured datasets with real-time enrichment. Describe what you want, get a live list.
The highest-accuracy web search for your AI. Fresh, fast, and affordable grounding in one call.
Get full or excerpted contents from the public web, including PDFs, JS-heavy pages, and government sites
Real-time access to the web's most valuable sources, built from the ground up for AIs.
Better, faster, cheaper at every price point
From sub-second lookups to hour-long research, state of the art at every tier
SimpleQA
| Provider | Accuracy (%) | Cost (CPM) |
|---|---|---|
| Parallel Turbo | 91 | 8 |
| Exa Instant | 89.3 | 20 |
| Tavily Ultra Fast | 72 | 23 |
| Brave Search | 87 | 16 |
| SerpAPI | 76.7 | 6 |
Coding
| Provider | Accuracy (%) | Cost (CPM) |
|---|---|---|
| Parallel Turbo | 79.7 | 131 |
| Exa Instant | 76.7 | 316 |
| Tavily Ultra Fast | 71.9 | 314 |
| Brave Search | 64.3 | 204 |
| SerpAPI | 54 | 131 |
| OpenAI Web Search | 76.7 | 475 |
BrowseComp
| Provider | Accuracy (%) | Cost (CPM) |
|---|---|---|
| Parallel Turbo | 51 | 216 |
| Exa Instant | 33.7 | 361 |
| Tavily Ultra Fast | 19.3 | 357 |
| Brave Search | 38.3 | 430 |
| SerpAPI | 23.3 | 999 |
FRAMES
| Provider | Accuracy (%) | Cost (CPM) |
|---|---|---|
| Parallel Turbo | 86 | 73 |
| Exa Instant | 88.3 | 133 |
| Tavily Ultra Fast | 80.7 | 194 |
| Brave Search | 81.3 | 146 |
| SerpAPI | 83.3 | 89 |
| OpenAI Web Search | 85.7 | 296 |
CPM: USD per 1000 requests. Cost is shown on a Log scale.
Dataset
We evaluated search providers against six open benchmarks covering complementary aspects of agentic search: BrowseComp (hard multi-hop questions that require navigating the live web), Frames (multi-document factoid reasoning), FreshQA (time-sensitive questions where the correct answer depends on recent web information), HLE (Humanity’s Last Exam — expert-level academic questions spanning math, science, and humanities), SealQA (ambiguity-robust factoid QA with intentionally misleading snippets), WebWalker (tasks designed around following links across pages to find an answer).
Evaluation methodology
Every task is run through a shared deep-research harness: a single GPT-5.4 agent is given two tools (web search and web fetch) with an iterative budget of up to MAX_TOOL_CALLS=25 tool calls per question. The agent plans sub-queries, fans out searches, fetches specific pages when snippets are insufficient, and returns an answer when it exhausts the number of allowed tool calls or has sufficient information to answer the question. Each answer is then LLM-graded by GPT-5.4. We report accuracy of the final answer.
We measure accuracy and overall cost, which includes LLM token costs and tool call costs.
Testing dates
April 19-21, 2026
Better, faster, cheaper at every price point
From sub-second lookups to hour-long research, state of the art at every tier
COST (CPM)
ACCURACY (%)
Loading chart...
CPM: USD per 1000 requests. Cost is shown on a Log scale.
Parallel
Others
Benchmark comparison across Cost (CPM) and Accuracy (%). CPM: USD per 1000 requests. Cost is shown on a Log scale.
Pioneers of the agentic web are building with Parallel
“Authoritative legal data across dozens of countries lives on sites no search engine has ever indexed. Parallel solves that at a scale we couldn't build ourselves.”
“The best agentic search isn't just the fastest or the cheapest. It's net new information that language models don't already know. Parallel delivers that.”
“Parallel provided us with a big upgrade in speed and index coverage. Combined with a 5x improvement in cost efficiency, it was an easy decision to build our agents with them.”
“You cannot make mistakes in financial services. With Parallel, our banking customers run agents that complete complex KYB checks in minutes, not hours.”
“In our internal evaluations we found Parallel to be the fastest and most accurate web search tool we tested. That speed together with its fine-grained control over web search means our agents can iteratively search, view results and refine search terms to get the answers users need.”
“We benchmarked providers on web research across entity types (companies, people, vessels), geographies, and data richness. Parallel topped every dimension on both breadth of findings and accuracy.”
“Parallel is core infrastructure for our agents. It outperformed every alternative on quality and cost, pulling from the whole web instead of prepackaged datasets.”
“Parallel is the highest accuracy API on the market. It handled the edge cases other providers simply couldn't: obscure companies, ambiguous names, conflicting sources.”
“Parallel's Monitor API lets Poke track anything our users care about: their team, their neighborhood, the news that matters to them. It helps make Poke more proactive in their daily lives, which ultimately makes for a better companion.”
“Parallel consistently delivered the highest quality results at the best price, and handled the complex government websites every other provider struggled with.”
“Parallel's team was incredibly helpful as we designed how Amp would use their APIs: tool names, params, UI, and other thoughtful guidance. They get it, and they care.”
“We chose Parallel as our preferred search API because of the comprehension and accuracy of their outputs: structured, well-sourced, and ready to feed straight into our content generation agents.”
“Our partnership with Parallel replaces repetitive human effort with continuous agentic research that integrates seamlessly into enhanced decision-making systems.”
“We tested every major web search provider. Parallel's outputs came back structured, well-cited, and dense with the right information.”
Pioneers of the agentic web are building with Parallel
Hear from the Pioneers[Hear from the Pioneers]”The best agentic search isn't just the fastest or the cheapest. It's net new information that language models don't already know. Parallel delivers that.”
”Parallel provided us with a big upgrade in speed and index coverage. Combined with a 5x improvement in cost efficiency, it was an easy decision to build our agents with them.”
”You cannot make mistakes in financial services. With Parallel, our banking customers run agents that complete complex KYB checks in minutes, not hours.”
”In our internal evaluations we found Parallel to be the fastest and most accurate web search tool we tested. That speed together with its fine-grained control over web search means our agents can iteratively search, view results and refine search terms to get the answers users need.”
”We benchmarked providers on web research across entity types (companies, people, vessels), geographies, and data richness. Parallel topped every dimension on both breadth of findings and accuracy.”
”Parallel is core infrastructure for our agents. It outperformed every alternative on quality and cost, pulling from the whole web instead of prepackaged datasets.”
”Parallel is the highest accuracy API on the market. It handled the edge cases other providers simply couldn't: obscure companies, ambiguous names, conflicting sources.”
”Parallel's Monitor API lets Poke track anything our users care about: their team, their neighborhood, the news that matters to them. It helps make Poke more proactive in their daily lives, which ultimately makes for a better companion.”
”Parallel consistently delivered the highest quality results at the best price, and handled the complex government websites every other provider struggled with.”
”Parallel's team was incredibly helpful as we designed how Amp would use their APIs: tool names, params, UI, and other thoughtful guidance. They get it, and they care.”
”We chose Parallel as our preferred search API because of the comprehension and accuracy of their outputs: structured, well-sourced, and ready to feed straight into our content generation agents.”
”Our partnership with Parallel replaces repetitive human effort with continuous agentic research that integrates seamlessly into enhanced decision-making systems.”
”We tested every major web search provider. Parallel's outputs came back structured, well-cited, and dense with the right information.”
Trust
Built for enterprise, secure by design
Build trust with the [Basis framework], unique to Parallel: calibrated confidence scores, citations, source excerpts, and reasoning traces.
- SOC 2 Type 2
- Zero data retention available
- HIPAA-ready offering
- GDPR Compliance
- Single sign-on (SSO/SAML)
Trust
Built for enterprise, secure by design
Build trust with the Basis framework, unique to Parallel: calibrated confidence scores, citations, source excerpts, and reasoning traces.
SOC 2 Type 2
Zero data retention available
HIPAA-ready offering
GDPR Compliance
Single sign-on (SSO/SAML)
Pay for answers, not tokens
Every Parallel API flexes to fit your needs. Dial speed, depth, and cost to what each workflow is worth, set a budget cap, and hold to it. Per-request pricing means you pay for answers, not tokens, so there’s no incentive to burn context and no surprise bill at the end of the month.
Start building for free
Get started with our APIs in seconds. Run up to 5,000 requests per month for free.
Agent onboarding prompt:
Use curl to read parallel.ai/agents.md and perform the setup to install Parallel
Available everywhere you build
- Google Cloud
- OpenRouter
- Vercel
- LangChain
- Supabase
- Google Sheets
- Snowflake
- Gemini
- Hermes Agent
- OpenClaw
- n8n
- MPP
Start building for free
Get started with our APIs in seconds. Run up to 5,000 requests per month for free.
Agent onboarding prompt
Use curl to read parallel.ai/agents.md and perform the setup to install Parallel
Available everywhere you build
FAQs
What does Parallel do?
Parallel builds web search and research APIs purpose-built for AI agents and agentic workflows. We run our own web-scale index with billions of pages and millions more added and updated daily. Our product suite spans the full range of knowledge work agents are helping automate for businesses: Search and Extract for real-time retrieval, Task and FindAll for deep research and entity discovery, and Monitor for continuous tracking. If your agent needs to read, research, or watch the web, Parallel is the unified platform that makes it as easy as an API call.
How is Parallel different from other web search APIs (like Exa, Tavily, or Perplexity)?
Most search APIs retro-fit traditional search engines for AI, but Parallel’s Index was designed for LLMs and programmatic use from the start. Our products are powered by innovations in crawling, indexing, and retrieval applied for efficient use of LLM context windows and software pipelines. By rebuilding the stack for the agentic software market, Parallel makes AI viable for high-stakes sectors like law, finance, and healthcare where accuracy is paramount.
Why does AI need the web?
Large language models are trained on past information, which means their knowledge of the present is out of date as soon as a training run begins. LLM knowledge cutoffs are often many months or even years behind present day facts. Web search APIs ground LLMs in real-world news published across the open (and in some cases, closed) web to ensure that AI always has the most up-to-date information needed to answer a query or task.
Which Parallel API should I use?
Start with the job you need done:
- Search API returns ranked URLs and dense excerpts in real time. Use it to ground an agent or assistant in fresh, cited web context or to find the most relevant pages (URLs).
- Extract API turns any public URL, including hosted PDFs and JavaScript-heavy pages, into clean markdown context. Use it to pull content from public URLs you already have or recently surfaced with a web search.
- Task API runs structured deep research and data enrichment with citations and confidence scoring. Use it for reports, account briefs, due diligence, or enriching records at scale.
- FindAll API discovers and structures entity datasets from a natural language query. Use it to build a live list of anything, including but not limited to events, people, and companies.
- Monitor API tracks the web continuously and sends webhook alerts when something appears or changes. Use it to watch prices, news, competitors, or regulations.
The general rule: Search and Extract when latency matters and the question is simple, Task and FindAll when you need synthesis and structure (LLM reasoning bundled), Monitor when the work never stops (ambient agents that proactively do work when triggered by new information).
How much does Parallel cost? Is there a free plan?
Parallel is pay-as-you-go, priced per request rather than per token, so you know what a call costs before it runs. Web search starts as low as $1 per 1,000 requests.
New accounts get a signup credit, and every account gets a recurring free monthly allowance of $5. Free credit is always spent before any paid balance, so you only pay once you’ve used it up. Parallel Search is also available via a free hosted MCP server, which is suitable for personal/hobbyist use.
See parallel.ai/pricing for the full breakdown. For higher rate limits and bespoke agreements, speak with our sales team.
What is the Parallel Web Index?
The Parallel Web Index is our proprietary, web-scale index of the open web, designed from the ground up to power the scale that AI web search demands. Our index contains billions of pages, and millions are added or updated daily, keeping it fresh and far-reaching.
What is Index by Parallel?
Index by Parallel is a platform that helps content owners understand how AI agents use their work and earn compensation tied to the value they contribute. Compensation is calculated by estimating each source’s Shapley value, its marginal contribution to the work an agent performs at the moment of inference. Content that’s uniquely valuable, hard to replace, or used in high-value agent work earns more.
How fresh is Parallel's data?
Parallel adds millions of pages to the Parallel Web Index daily, and you can force a live crawl for time-sensitive queries. Set freshness controls to require recent pages or trigger a fresh fetch, so your agent always reads what’s true today.
How does Parallel cite its sources?
Search and Extract APIs include source URLs.
On Parallel’s agentic APIs (Web Agents), every output supports Basis, our verifiability framework, which attaches citations, the reasoning behind a result, and a calibrated confidence score to each fact. You can trust an answer in production and audit it afterward, tracing any claim back to the page it came from. Those confidence scores also feed back into retrieval, so results improve over time.
Is Parallel secure and compliant?
Yes. Parallel is SOC 2 Type II certified, HIPAA compliant, and offers zero data retention. For teams with stricter requirements, enterprise controls are available. See our terms of service for more information.
How do I connect Parallel to my agent or existing stack?
You have a few options. Call the REST API directly, or use the Python (parallel-web) or TypeScript SDK. For agent frameworks, the Parallel Search MCP server drops in through Cursor, Claude Code, and other MCP-compatible tools. If you already use the OpenAI SDK, point your base URL at the Chat API and swap your key; everything else works the same. Parallel also integrates with LangChain, the Vercel AI SDK, and Google Vertex AI, and delivers async results over webhooks. Start at docs.parallel.ai.
FAQs
Parallel builds web search and research APIs purpose-built for AI agents and agentic workflows. We run our own web-scale index with billions of pages and millions more added and updated daily. Our product suite spans the full range of knowledge work agents are helping automate for businesses: Search and Extract for real-time retrieval, Task and FindAll for deep research and entity discovery, and Monitor for continuous tracking. If your agent needs to read, research, or watch the web, Parallel is the unified platform that makes it as easy as an API call.
Most search APIs retro-fit traditional search engines for AI, but Parallel’s Index was designed for LLMs and programmatic use from the start. Our products are powered by innovations in crawling, indexing, and retrieval applied for efficient use of LLM context windows and software pipelines. By rebuilding the stack for the agentic software market, Parallel makes AI viable for high-stakes sectors like law, finance, and healthcare where accuracy is paramount.
Large language models are trained on past information, which means their knowledge of the present is out of date as soon as a training run begins. LLM knowledge cutoffs are often many months or even years behind present day facts. Web search APIs ground LLMs in real-world news published across the open (and in some cases, closed) web to ensure that AI always has the most up-to-date information needed to answer a query or task.
Start with the job you need done:
- Search API returns ranked URLs and dense excerpts in real time. Use it to ground an agent or assistant in fresh, cited web context or to find the most relevant pages (URLs).
- Extract API turns any public URL, including hosted PDFs and JavaScript-heavy pages, into clean markdown context. Use it to pull content from public URLs you already have or recently surfaced with a web search.
- Task API runs structured deep research and data enrichment with citations and confidence scoring. Use it for reports, account briefs, due diligence, or enriching records at scale.
- FindAll API discovers and structures entity datasets from a natural language query. Use it to build a live list of anything, including but not limited to events, people, and companies.
- Monitor API tracks the web continuously and sends webhook alerts when something appears or changes. Use it to watch prices, news, competitors, or regulations.
The general rule: Search and Extract when latency matters and the question is simple, Task and FindAll when you need synthesis and structure (LLM reasoning bundled), Monitor when the work never stops (ambient agents that proactively do work when triggered by new information).
Parallel is pay-as-you-go, priced per request rather than per token, so you know what a call costs before it runs. Web search starts as low as $1 per 1,000 requests.
New accounts get a signup credit, and every account gets a recurring free monthly allowance of $5. Free credit is always spent before any paid balance, so you only pay once you’ve used it up. Parallel Search is also available via a free hosted MCP server, which is suitable for personal/hobbyist use.
See parallel.ai/pricing for the full breakdown. For higher rate limits and bespoke agreements, speak with our sales team.
The Parallel Web Index is our proprietary, web-scale index of the open web, designed from the ground up to power the scale that AI web search demands. Our index contains billions of pages, and millions are added or updated daily, keeping it fresh and far-reaching.
Index by Parallel is a platform that helps content owners understand how AI agents use their work and earn compensation tied to the value they contribute. Compensation is calculated by estimating each source’s Shapley value, its marginal contribution to the work an agent performs at the moment of inference. Content that’s uniquely valuable, hard to replace, or used in high-value agent work earns more.
Parallel adds millions of pages to the Parallel Web Index daily, and you can force a live crawl for time-sensitive queries. Set freshness controls to require recent pages or trigger a fresh fetch, so your agent always reads what’s true today.
Search and Extract APIs include source URLs.
On Parallel’s agentic APIs (Web Agents), every output supports Basis, our verifiability framework, which attaches citations, the reasoning behind a result, and a calibrated confidence score to each fact. You can trust an answer in production and audit it afterward, tracing any claim back to the page it came from. Those confidence scores also feed back into retrieval, so results improve over time.
Yes. Parallel is SOC 2 Type II certified, HIPAA compliant, and offers zero data retention. For teams with stricter requirements, enterprise controls are available. See our terms of service for more information.
You have a few options. Call the REST API directly, or use the Python (parallel-web) or TypeScript SDK. For agent frameworks, the Parallel Search MCP server drops in through Cursor, Claude Code, and other MCP-compatible tools. If you already use the OpenAI SDK, point your base URL at the Chat API and swap your key; everything else works the same. Parallel also integrates with LangChain, the Vercel AI SDK, and Google Vertex AI, and delivers async results over webhooks. Start at docs.parallel.ai.