The web access layer for agents
Tavily is a search-and-retrieval API purpose-built for AI agents and RAG pipelines, offering search, extraction, crawling, site mapping, and a higher-level "research task" endpoint that returns LLM-ready results rather than raw SERPs.
Published by the vendor themselves — not community forks.
Production-ready MCP server exposing real-time search, extract, map, and crawl
Agent skill repository published by Tavily
| Plan | Price | Included | Notable |
|---|---|---|---|
| Researcher (Free) | $0/mo | 1,000 API credits/month | No credit card required, Email support |
| Pay As You Go | $0.008/credit | Usage-based, no monthly allotment | Cancel anytime, Email support |
| Project | Adjustable monthly rate (usage slider) | 4,000 API credits/month (base) | Higher rate limits, Email support |
| Enterprise | Contact for pricing | Custom API call volume | Custom rate limits, SLAs, Enterprise-grade security and privacy |
Checked 2026-07-31 against their pricing page. Vendors change pricing without notice.
It bills per API credit, with different endpoints and depth settings consuming different credit amounts, and exposes the same capabilities through a hosted remote MCP server for AI-assistant clients (Claude Desktop, Claude Code, Cursor, OpenAI clients).
Developers building AI agents, RAG pipelines, or research/automation tools who need live web search and content extraction as an API primitive rather than a consumer search product; the vendor names MongoDB, LangChain, JetBrains, AWS, Mastercard, and IBM as users.
Credit pricing is granular and stacks per capability - advanced search costs 2x basic, crawl combines separate mapping and extraction charges, and the Research Task endpoint alone can cost 15-250 credits per call on the pro model - so agentic workflows that chain several of these calls per task can consume the free 1,000-credit/month tier very quickly. No OpenAPI or machine-readable spec was found on the docs site, so teams that want to generate a typed client automatically will need to hand-write one against the REST reference instead.
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
No machine-readable OpenAPI/Swagger spec was found on the docs site at check time; the API reference is presented as per-endpoint documentation pages instead.
Yes, the Researcher plan gives 1,000 API credits/month at no cost, with no credit card required.
Monthly credit-allotment plans (Free through Project, with an adjustable usage slider) or pay-as-you-go at $0.008/credit, plus custom Enterprise pricing; credit cost varies by endpoint and depth (e.g. advanced search costs 2x basic search).
Dynamic pricing depending on query complexity: 4-110 credits on the mini model, 15-250 credits on the pro model.
Crawl combines the cost of mapping plus extraction on the pages it touches - e.g. crawling 10 pages with basic extraction costs 1 credit (mapping) + 2 credits (extraction) = 3 credits total.
This page tells you whether Tavily 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.