Tavily
Real-time web access for AI agents
- Tool type
- Web search and research APIs for AI
- Made for
- Agent and application developers
- Runs on
- APIs, SDKs and MCP connections
What it does
Tavily is an API that gives AI agents live web access. It takes search queries, URLs, sites to crawl, or research questions, and returns clean, structured, chunked content that models can reason over. Outputs include search results, extracted page content, crawled site pages, site maps, and cited research reports.
Developers typically call Tavily from code through its Python or JavaScript SDKs, or connect coding agents such as Claude, Cursor, ChatGPT and Codex via MCP. Requests pass through built-in security and privacy checks that can block PII in queries and prompt injection. The docs also describe keyless access for Search and Extract without an API key.
Tavily suits developers building agents or grounded AI applications, and students who qualify for the free student offer. Access requires a free account and API key for most features, and usage is metered in API credits. Limitations: credits are consumed by request type, and requests stop when monthly credits run out on the free tier.
What it can do
- Real-time web search returning structured, chunked results for models
- Page extraction that pulls clean content from one or more URLs
- Site crawling to read many pages across a site
- Site mapping to discover a site's structure
- Cited multi-source research reports
- PII and prompt-injection safeguards on requests
Works with
- Claude and Claude Code (MCP)
- ChatGPT and Codex (MCP)
- Cursor (MCP)
- OpenCode plugin
- LangChain
- OpenAI
- Vercel AI SDK
- n8n
- CrewAI
- Mastra
Example workflows
- Grounding a research agent with live search
- Illustrative example. Input: a user question about a company. Steps: call Search with the query, receive structured chunks, pass them to the model as context. Output: an answer grounded in retrieved web sources.
- Extracting content from a list of URLs
- Illustrative example. Input: a set of article URLs. Steps: call Extract with the URLs, receive cleaned page content, feed it into a summarization step. Output: clean text per URL ready for downstream processing.
Before you choose
- API credits are consumed based on request type, with heavier requests using more credits; detailed usage is in the documentation.
- On the free plan, requests stop when monthly credits run out until they reset on the first of the month or you upgrade.
- Some features are reached through an account, and Keyless access covers only Search and Extract.
Listed under
Get started
Free plan includes 1,000 API credits per month with no credit card required.
- Sign up for a free Tavily account and copy your API key from the dashboard.
- Install the Python (pip install tavily-python) or JavaScript (npm i @tavily/core) SDK and set your API key.
- Make a first Search call, or add the MCP server to a supported coding agent such as Claude Code or Cursor.
Plans and prices
Prices in USDResearcher includes 1,000 credits per month, resetting on the first day of each month. Pay As You Go costs $0.008/credit. The rendered Project entry tier is $30/month for 4,000 credits. Credits are not requests: the operation and depth determine consumption. Enterprise terms are quoted.
- 1,000 credits/month
- No card required
- 4,000 credits/month
Source: tavily.com/pricing, checked 29 Oct 2026. Prices can change; check before you buy.
Sources for this overview
Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.
- Tavily ↗www.tavily.com
- Tavily Pricing ↗www.tavily.com
- Connect your AI agent to the web - Tavily Docs ↗docs.tavily.com
- NEBIUS GROUP N.V._December 31, 2025 ↗www.sec.gov
Ratings and reviews
No ratings for Tavily yet.
What’s changed
Free accountSignals
Free accountData and security
Free accountCompare with similar tools
Free accountWhere this comes from
How we research- Company website
- tavily.com: product, pricing and policy pages
- YC directory mirror
- Batch and company name
- Social accounts
- Accounts the company links, and its posts on X
- Human review
- Checked by a person
- knowledge_base
Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.