Tavily Search API

Tavily is a search API built specifically for AI agents and LLM applications. Unlike general-purpose search APIs, Tavily returns clean, structured results optimized for LLM context injection — full article content, not just snippets. It handles query optimization, result filtering, and context compression internally. Supports both search and topic-specific retrieval (news, general). Has an official LangChain integration and MCP server.

Evaluated Mar 06, 2026 (0d ago) vcurrent

Homepage ↗ Repo ↗ Developer Tools search web-search ai rag agent-tools research mcp

⚙ Agent Friendliness

/ 100

Can an agent use this?

🔒 Security

/ 100

Is it safe for agents?

⚡ Reliability

/ 100

Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality

Documentation

Error Messages

Auth Simplicity

Rate Limits

🔒 Security

TLS Enforcement

100

Auth Strength

Scope Granularity

Dep. Hygiene

Secret Handling

Simple API key auth. No scope granularity — key has full search access. Young platform with growing security posture. Search results include external content — agents should validate sources.

⚡ Reliability

Uptime/SLA

Version Stability

Breaking Changes

Error Recovery

Best When

Agent needs to find and retrieve current web information as part of a reasoning loop — Tavily returns ready-to-use content, not just links.

Avoid When

You need to scrape a specific known URL — use Firecrawl or Jina Reader instead.

Use Cases

• Web research in ReAct-style agents (search → reason → search loop)
• Grounding LLM responses with current web information
• News monitoring and event tracking for agents
• Building RAG systems that need fresh web content
• Replacing manual browsing in research workflows

Not For

• Image search or multimedia retrieval
• High-volume scraping (rate limits apply)
• Private/intranet content that isn't indexed publicly

Interface

REST API

Yes

GraphQL

gRPC

MCP Server

Yes ↗

SDK

Yes

Webhooks

Authentication

Methods: api_key

OAuth: No Scopes: No

Single API key. No scopes. Pass as tvly-api-key header or in request body. Very simple — designed for quick agent integration.

Pricing

Model: freemium

Free tier: Yes

Requires CC: No

Credits are consumed per search. Results depth setting (basic vs advanced) affects credit consumption. Very competitive pricing for AI agent use cases.

Agent Metadata

Pagination

none

Idempotent

Full

Retry Guidance

Documented

Known Gotchas

⚠ Results are time-sensitive — the same query returns different results as the web changes, so don't cache for long
⚠ include_raw_content: true is expensive in tokens — use sparingly in context-limited agents
⚠ max_results defaults to 5; increasing to 10+ increases token cost significantly
⚠ Search topic 'news' returns recent articles but may miss older authoritative sources
⚠ No way to filter by date range — all results are ranked by Tavily's internal relevance
⚠ Rate limits are per API key, so shared keys in multi-agent systems deplete quickly

Alternatives

{'id': 'firecrawl-api', 'reason': 'Firecrawl is better when you have a specific URL to read; Tavily is better for open-ended search queries'} {'id': 'jina-ai-api', 'reason': 'Jina Reader is free for URL-to-markdown; Tavily wins for query-based search with relevance ranking'}

Full Evaluation Report

Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for Tavily Search API.

$99

API endpoint ↗ Agent guide ↗ Report inaccuracy

Scores are editorial opinions as of 2026-03-06.