Where agents find answers
Parallel is web infrastructure built for AI agents rather than humans: a family of APIs (Search, Extract, Task, FindAll, Monitor, Responses/Chat) that let agents search the live web, pull clean markdown from any URL, run cited multi-hop research, discover and structure entity datasets, and get webhook alerts when tracked pages change.
Published by the vendor themselves — not community forks.
Open-source Agent Skills for web search, extraction, deep research, data enrichment, and entity discovery, installable via npx skills / the parallel-cli into Cursor, Cline, GitHub Copilot, Claude Code, and 30+ other coding agents.
Rebuilding web search for AI agents | Parallel Web Systems — Parallel Web Systems
| Plan | Price | Included | Notable |
|---|---|---|---|
| Free | $0 | Up to 5,000 requests/month; up to $80 signup credit plus $5 monthly free credits | Search API, Extract API, Task API access at pay-as-you-go rates |
| Pay-as-you-go | Task API: $0.005-$2.40/request; Search API: $0.001-$0.005 per 10 results; Extract API: $0.001/request; Responses API: $0.01-$0.25/request; Monitor API: $0.003-$0.01/request; FindAll API: $0.03-$1.00/match | — | All API products, Tiered processors per product (e.g. Task API: Lite $5, Base $10, Core $25, Core2x $50, Pro $100, Ultra $300, Ultra2x $600, Ultra4x $1,200, Ultra8x $2,400 per 1,000 requests) |
| Enterprise | Contact for pricing | — | Zero Data Retention, Data Protection Agreements, Single Sign-On |
Checked 2026-07-31 against their pricing page. Vendors change pricing without notice.
Parallel (Parallel Web Systems) builds a set of task-specific web APIs aimed squarely at AI agents rather than at rendering pages for humans: Search and Extract for grounding, Task for cited deep research and structured enrichment, FindAll for entity discovery, and Monitor for continuous tracking.
Teams building AI agents or AI-powered products that need verified, cited, real-time web data as a primary input - GTM/enrichment tooling, research and due-diligence products, and vertical AI agents in regulated or high-stakes domains (the vendor names law, finance, and healthcare). The vendor lists Notion, Dropbox, and Harvey as customers.
Pricing is metered per request/match and varies widely by processor tier (Task API alone ranges from $5 to $2,400 per 1,000 requests depending on tier), so teams wanting a flat, predictable bill will find costs hard to forecast without testing tiers first. This is API infrastructure, not a UI product - teams that want a point-and-click research or enrichment tool without engineering integration work are better served elsewhere. FindAll and Monitor also charge separately from Search/Extract, so a workflow that touches several of Parallel's APIs stacks multiple metered line items.
APIs that answer a query with retrieved web content, built to be called by an agent rather than read by a person.
Category: Search and research APIs. Membership is decided by 2 stated tests, not by category tags — see them and the full comparison.
Appears in comparison shortlists
Metered pay-as-you-go per API: Task API $0.005-$2.40/request across nine processor tiers (Lite through Ultra8x), Search $0.001-$0.005 per 10 results, Extract $0.001/request, Responses $0.01-$0.25/request, Monitor $0.003-$0.01/request, FindAll $0.03-$1.00/match. A free tier covers up to 5,000 requests/month.
Yes - up to 5,000 requests/month free, plus up to $80 signup credit and $5/month ongoing free credit.
A structured result matching a JSON schema the caller defines, with citations and confidence scoring, from a multi-hop research process; results can be retrieved via blocking call, polling, or server-sent events.
Yes - parallel-agent-skills on the vendor's GitHub org, an MIT-licensed collection of Agent Skills for web search, extraction, research, and enrichment, installable into 30+ coding agents via npx skills.
Native SQL-native functions/UDFs for Snowflake, BigQuery, and Apache Spark, plus integrations for DuckDB, Polars DataFrames, Supabase Edge Functions, and a Google Sheets PARALLEL_QUERY function.
This page tells you whether Parallel fits. Cheetah builds and runs the system it goes into — enrichment, sequencing, CRM and the joins between them.
Book a call →The directory stays free. There is nothing gated behind this.