Parallel
Web search and research APIs built for AI agents, priced per request
Overview
Parallel (Parallel Web Systems) offers web APIs for AI agents — Search, Extract, Task for multi-step cited research, an OpenAI-compatible Responses API, Monitor and FindAll — priced per request from $1 per 1,000 searches. Founded by Parag Agrawal; SOC 2 Type II, HIPAA and zero data retention.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 1 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- +$1 per 1K additional results
- 200ms-3s latency
- 600 requests/min
- Cached 1-3s, live 60-90s
- Structured, cited research fields
- Lite 5-60s, Ultra8x 8-30 min
- OpenAI Responses-compatible
- Hourly, daily or weekly monitoring
- Entity list building
- Zero data retention, DPA, SSO
- Custom rate limits
What you can produce with Parallel
- Search API with ranked URLs and excerpts
- Task API for multi-step, cited research (Lite to Ultra8x)
- Extract, Responses, Monitor and FindAll APIs
- Data Connectors (PubMed, ClinicalTrials.gov, bring-your-own licences)
- Per-request pricing from $1 per 1,000
- SOC 2 Type II, HIPAA and zero data retention
- MCP server, CLI, Python and TypeScript SDKs
Parallel is one of the strongest web-research APIs for agent builders: fast and cheap search from $1 per 1,000 requests, cleaner extraction, and a Task API that returns structured, cited answers at research depths up to half-hour runs. Enterprise basics — SOC 2 Type II, HIPAA, zero data retention — are in place, it is well funded after a $2 billion Series B, and Google has built it into Gemini Enterprise. Treat its benchmark claims carefully: its "#1" announcement in August has been overtaken on the independent Artificial Analysis leaderboard by Perplexity and others. Costs climb steeply at the deepest Task levels, so match the processor to the question. The free monthly credit covers small projects. A strong default for agents that need grounded web data; compare with Exa, Tavily and Perplexity's API on your queries.
What we found
Parallel, from Parallel Web Systems, builds web-research APIs for AI agents rather than for people. Founded in 2023 by Parag Agrawal, formerly chief executive of Twitter, and based in San Francisco, it sells a family of endpoints priced per request: Search returns ranked web results with relevant excerpts; Extract pulls clean content from pages; the Task API runs multi-step research at processor levels from Lite to Ultra8x, returning structured, cited fields; the OpenAI Responses-compatible Responses API answers questions with web grounding; Monitor watches the web for changes; and FindAll builds lists of entities matching criteria. Data Connectors add sources such as PubMed and ClinicalTrials.gov, or licensed databases like Crunchbase if you bring your own access. It is SOC 2 Type II certified, HIPAA compliant and offers zero data retention.
Growth has been fast. Parallel raised a $100 million Series A co-led by Kleiner Perkins and Index Ventures in late 2025 and a $100 million Series B led by Sequoia at a $2 billion valuation in April 2026, bringing its total to $230 million; it says more than 100,000 developers use it, and Harvey, Notion and Opendoor appear among customers. In July it partnered with Google Cloud to put Parallel search inside Gemini Enterprise. In May it launched Index, a platform intended to pay publishers whose content agents use.
On quality, Parallel announced in August that it ranked first on Artificial Analysis's search index; the independent leaderboard's September update places its advanced search behind two Perplexity configurations and Octen, level with Brave. Any search API should be tested on your own queries.
Pricing is per 1,000 requests: Search costs $1 in its fast modes and $5 in its basic and advanced modes for ten results; Extract $1; Task from $5 at Lite to $2,400 at Ultra8x; Responses $10 to $250; FindAll a base fee plus a charge per match. Accounts with a card on file get $5 of free credit a month, about 5,000 requests, plus up to $80 at sign-up.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including Parallel's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Parallel unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Parallel directly →
Spotted something out of date? Suggest an update →
More in Research & Data
