API review

Tavily API review

Credit-based web search, extract, map, crawl and research API built for AI agents, with a hosted MCP server and a keyless free mode. Best for agent web search with LLM-ready snippets at 1-2 credits per query.

Verified 25 Oct 2026 · 10 endpoints listed · Official MCP server

See pricing

Tavily API docs

What a call costs

The actions you pay for, as Tavily prices them in credits.

Action	Cost	Per
Search basic / fast / ultra-fast	1 credit	request
Search advanced	2 credits	request
Extract basic	1 credit	5 successful URLs
Extract advanced	2 credits	5 successful URLs
Map	1 credit (2 with instructions)	10 successful pages
Crawl	map cost + extract cost	request
Research mini	4-110 credits (measured: 7 credits for one mini run, derived from dashboard total)	request
Research pro	15-250 credits	request

Show all 10 priced actionsHide the rest

Action	Cost	Per
Startup program	100k free credits	one-off
Free for students	free	student programme (terms on pricing page)

Pricing, plan by plan

US dollars, as Tavily's pricing page shows them on 25 Oct 2026. Monthly and annual billing are shown apart.

Free tier Researcher: 1,000 credits/month, no card; credits reset on the 1st of each month

Researcher

Free

1,000 credits/month, no card; dev keys only (Production key option disabled on free); Usage & logs dashboard is paid-only

API included

Plan details

Requests stop until reset unless PAYGO enabled

Project

$30/mo

billed monthly

$25/mo

billed annually

4,000 credits/month ($0.0075/credit)

API included

Plan details

PAYGO $0.008/credit if enabled

Bootstrap

$100/mo

billed monthly

$83.33/mo

billed annually

15,000 credits/month ($0.0067/credit)

API included

Plan details

PAYGO $0.008/credit if enabled

Startup

$220/mo

billed monthly

$183.33/mo

billed annually

38,000 credits/month ($0.0058/credit)

API included

Plan details

PAYGO $0.008/credit if enabled

Growth

$500/mo

billed monthly

$416.67/mo

billed annually

100,000 credits/month ($0.005/credit)

API included

Plan details

PAYGO $0.008/credit if enabled

Pay as you go

One-off pack

$0.008 per credit, billed incrementally once plan credits run out; enables Production keys

API included

Show all 7 plansShow fewer plans

Enterprise

Custom quote

Custom

API included

Rate limits

Per key environment: Development 100 RPM, Production 1,000 RPM; /crawl 100 RPM both; /research 20 RPM to create; /usage 10 per 10 min.

Limit	Development key (all plans)	Production key (paid plan or PAYGO only; disabled on free in-app)	Keyless (X-Tavily-Access-Mode: keyless; /search and /extract only)
Limit	100 RPM; /crawl 100 RPM; /research create 20 RPM; /usage 10 per 10 min	1,000 RPM; /crawl still 100 RPM; /research create 20 RPM	free, rate-limited, cap not published; different error format

How to get an API key

Tavily uses API key. The steps below follow Tavily's own docs.

Create a Tavily account. The Researcher plan includes API access.
Sign up at app.tavily.com (free, no card; Auth0 login with Cloudflare Turnstile). A 'default' dev key exists at signup; Production keys need a paid plan or PAYGO; every key carries a 'Limit monthly usage' credit cap. /org-usage needs the organization owner's PERSONAL key.
Send it with every request: header: Authorization: Bearer tvly-<key> (spec bearerAuth); keyless via header X-Tavily-Access-Mode: keyless (/search, /extract only; a Bearer key takes precedence); optional X-Project-ID header for project attribution.
Make a first call: Web search returning ranked, LLM-ready results; optional answer, raw content, images. Keep the key in an environment variable, never in the command.
```
curl -X POST "https://api.tavily.com/search" \
  -H "Authorization: Bearer $TAVILY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{}'
```

Where the docs live

What	Where
Documentation	docs.tavily.com
API reference	docs.tavily.com/documentation/api-reference/endpoint/search
OpenAPI spec	docs.tavily.com/documentation/api-reference/openapi.json
llms.txt	docs.tavily.com/llms.txt
Changelog	docs.tavily.com/changelog
Base URL	`https://api.tavily.com`
MCP server	`https://mcp.tavily.com/mcp/`
SDKs	tavily-python, @tavily/core (JavaScript), tavily-agent-toolkit (Python), tavily-cli (pip install tavily-cli; command tvly; installed by the dashboard setup prompt tavily.com/agent-setup/SKILL.md), Vercel AI SDK v5 integration, LangChain / LlamaIndex integrations

Strengths and limits

Where it fits

Agent web search with LLM-ready snippets at 1-2 credits per query
Bulk URL text extraction (5 URLs per credit on basic)
Quick site maps and shallow crawls without running a scraper
Prototyping with zero setup (keyless header, no account)

Where it falls short

Hard anti-bot targets or logged-in sites (no proxy/unblocker product)
Entity-level company/people search
Publishing benchmark comparisons (terms forbid disclosing performance analysis)

Worth knowing

Credits reset on the 1st of each calendar month regardless of billing date; unused plan credits do not roll over.
When monthly credits run out, requests stop unless pay-as-you-go is enabled.
Production keys (1,000 RPM) require a paid plan or PAYGO; dev keys are 100 RPM.
Advanced search costs 2x; advanced extract costs 2x; map with instructions costs 2x.
Research is dynamically priced, up to 250 credits ($1.25-$2.00) per pro request.

10 moreShow fewer

Terms (updated 4 May 2026) forbid disclosing performance information or analysis to third parties, building a competing product, and automated high-risk decisions without human oversight.
Crawl has its own 100 RPM cap even on production keys.
GET /usage did not reflect 14 used credits on the free plan (returned 0 twice while the dashboard showed 14); do not rely on it for live budget checks.
Usage & logs dashboard is paid-only on the free Researcher plan.
Settings > 'Allow use of query data' is ON by default ('We may use queries to improve the quality of our services'); turn it off for client work.
Credits are counted per started block and look rounded down: 1 extracted URL = 0 credits, 21 mapped URLs = 2, 5 crawled pages = 1.
/extract accepts at most 20 URLs per request.
Keyless mode covers /search and /extract only; /map, /crawl and /research need a key.
EU endpoint api.eu.tavily.com / mcp.eu.tavily.com is for selected customer evaluations only, enabled per key.
App and login screens are branded 'tavily by NEBIUS'; the contracting entity in the Terms is still AlphaAI Technologies Inc. dba Tavily.

Who should use it

Use Tavily for: Agent web search with LLM-ready snippets at 1-2 credits per query
Look elsewhere for: Hard anti-bot targets or logged-in sites (no proxy/unblocker product)

If Tavily is not the answer

Other Research and scraping tools in this directory, each with its own review.

DataForSEO

Research and scraping

Best for cheap bulk SERP pulls via the Standard queue.

DataForSEO API review

Exa

Research and scraping

Free planOfficial MCP

Best for agent web search and RAG grounding with token-efficient contents.

Exa API review

Parallel AI

Research and scraping

Free planOfficial MCP

Best for row-level account/contact research and enrichment with per-field citations and confidence.

Parallel AI API review

Compare Tavily with

Exa side by side

Before you build

Four things to settle first, each read off Tavily's own pages.

Confirm the API is on your planThe Researcher plan includes API access. Free tier: Researcher: 1,000 credits/month, no card.
Keep the key out of your codeAPI key. It is sent as: header: Authorization: Bearer tvly-<key> (spec bearerAuth); keyless via header X-Tavily-Access-Mode: keyless (/search, /extract only; a Bearer key takes precedence); optional X-Project-ID header for project attribution.
Design around the rate limitPer key environment: Development 100 RPM, Production 1,000 RPM; /crawl 100 RPM both; /research 20 RPM to create; /usage 10 per 10 min.
Know the catch before you scaleCredits reset on the 1st of each calendar month regardless of billing date; unused plan credits do not roll over.

Want us to build it for you?

Tell us what you want Tavily to do in your stack. We will show you how we would wire it into your data, outbound or CRM, and what we would check first.

Book a call

See case studies

Read from docs.tavily.com/documentation/api-credits, docs.tavily.com/documentation/api-reference/endpoint/search and docs.tavily.com/documentation/rate-limits. Verified 25 Oct 2026.

All 10 endpointsFor developers: method, path, what it does and what it costs.

Base URL: api.tavily.com

Bulk and async: /extract takes 1-20 URLs per request (spec); /crawl and /research run server-side; no separate batch API

Filter endpoints

10 endpoints

Method	Path	What it does	Cost
POST	`/search`	Web search returning ranked, LLM-ready results; optional answer, raw content, images	basic/fast/ultra-fast 1 credit, advanced 2 credits per request
POST	`/extract`	Clean content from 1-20 URLs per request	basic: 1 credit per 5 successful URLs; advanced: 2 credits per 5; failed URLs free; measured: 1 URL basic = 0 credits
POST	`/crawl`	Crawl a site from a base URL (map + extract)	map cost + extract cost, e.g. 10 pages basic = 3 credits, advanced = 5 credits
POST	`/map`	Discover a site's URLs from a base URL	1 credit per 10 successful pages; 2 per 10 with instructions; failed requests free
POSTasync	`/research`	Start an async multi-source research task	dynamic: model=mini 4-110 credits, model=pro 15-250 credits per request; measured one mini run = 7 credits
GETasync	`/research/{request_id}`	Poll research task status and result	Not stated
POST	`/feedback`	Submit feedback on a search request or session	Not stated
GET	`/usage`	API key and account usage	free
POST	`/logs`	Usage logs per API key (filter by key, endpoint)	Not stated
POST	`/org-usage`	Organization usage across all keys (needs org owner's personal key)	Not stated