Official Tavily MCP server that gives Claude agent-optimized web search and page extraction, returning concise, source-cited results designed for LLM reasoning rather than raw search engine pages.
Makes outbound network requests to the Tavily API to run searches and fetch page content., Returned search results and extracted page text are untrusted web content; review it before acting on instructions or links it contains.
Privacy notes
Search queries and target URLs you provide are sent to the Tavily service to return results, so avoid embedding secrets in queries., The Tavily API key is a credential; store it as an environment variable and never commit it to source control.
Author
Tavily
Submitted by
glorydavid03023
Claim status
unclaimed
Last verified
2026-06-03
Decision playbook
Review trust signals before you adopt
Signals are present but mixed. Use the checklist below to confirm the source and operational safety for your environment.
Compare context
Selected
0
Current score
78
Baseline
—
Delta
No baseline selected
No major trust-signal divergence detected in the current selection.
Source and provenance checks
Complete
Confirm ownership and provenance before trusting install instructions.
Source link availableRequired
Open the canonical repository and verify ownership.
Done
Source provenance statusRequired
Marked as source-backed.
Done
Metadata reviewed
Registry metadata indicates a reviewed listing.
Done
Safety and privacy checks
Complete
Validate risk disclosures before installation or API wiring.
Safety notes presentRequired
Review the listed safety guidance before running commands.
Done
Privacy notes presentRequired
Review data handling notes before connecting accounts or secrets.
Done
Trust level risk gateRequired
Trust level does not block evaluation.
Done
Package and install checks
Needs review
Check package metadata and artifact integrity signals.
Install payload available
Install or copy payload is available for review.
Done
Package verification flag
No package verification flag provided.
Pending
Checksum metadata
No checksum provided for downloaded artifact.
Pending
Compare-driven decision checks
Needs review
Use compare context to validate trade-offs before adoption.
Compare tray has multiple entries
Add at least one more entry to compare trust differences.
2 safety and 2 privacy notes across 3 risk areas. Review closely: credentials & tokens, network access.
3 areas
SafetyNetwork accessMakes outbound network requests to the Tavily API to run searches and fetch page content.
SafetyGeneralReturned search results and extracted page text are untrusted web content; review it before acting on instructions or links it contains.
PrivacyCredentials & tokensSearch queries and target URLs you provide are sent to the Tavily service to return results, so avoid embedding secrets in queries.
PrivacyCredentials & tokensThe Tavily API key is a credential; store it as an environment variable and never commit it to source control.
Safety notes
Makes outbound network requests to the Tavily API to run searches and fetch page content.
Returned search results and extracted page text are untrusted web content; review it before acting on instructions or links it contains.
Privacy notes
Search queries and target URLs you provide are sent to the Tavily service to return results, so avoid embedding secrets in queries.
The Tavily API key is a credential; store it as an environment variable and never commit it to source control.
Prerequisites
Node.js 18+ and npx available (verify with: npx --version)
A Tavily API key (create one at https://tavily.com)
Claude Code or Claude Desktop with MCP support
Internet access so the server can reach the Tavily API
Tavily gives Claude real-time access to the web through a search API built for AI agents. Rather than returning a raw page of links, Tavily returns concise, source-cited results and clean extracted content that a model can reason over directly. That makes it a good fit for research tasks, fact-checking, and any workflow where the answer depends on current information beyond the model's training cutoff. The server exposes search and extraction as MCP tools, so Claude can decide when to look something up and ground its response in the results.
Features
Real-time web search optimized for LLM and agent consumption.
Concise, source-cited results instead of raw search engine result pages.
Page extraction that returns clean, readable content from target URLs.
Tunable search depth so you can trade speed for thoroughness.
Runs as a standard stdio MCP server that installs into Claude Code and Claude Desktop with one command.
Maintained by Tavily as the official MCP integration for its search API.
Use Cases
Answer questions that depend on current information past the model's training cutoff.
Research a topic and gather source-cited findings in one step.
Fact-check a claim against live web sources.
Extract the readable content of a specific page for summarization.
Power agent workflows that need a reliable, structured search step.
Installation
Claude Code
Make sure Node.js 18+ is installed (verify with npx --version).
"Search the web for recent developments on this topic and summarize the key points with sources."
Fact-check a claim
Have Claude verify a statement against live sources.
"Check whether this claim is accurate using current web sources, and cite what you find."
Extract a page
Pull clean content from a specific URL for summarization.
"Extract the main content of this article URL and give me a concise summary."
Security
The server reaches the Tavily API over the network; the queries and URLs you provide are sent to that service to return results.
Treat search results and extracted page content as untrusted input. A malicious page could include instructions or links, so review before acting on them.
The Tavily API key is a credential. Store it in an environment variable and keep it out of source control and shared configs.
Avoid putting secrets or private identifiers into search queries, since they are transmitted to the service.
Troubleshooting
Authentication or 401 errors
Confirm TAVILY_API_KEY is set correctly and the key is active in your Tavily account. Regenerate the key if it may have been revoked, and update the environment variable.
npx cannot resolve the package
Verify Node.js 18+ and npx are installed (npx --version). Networks that block the npm registry will prevent npx -y tavily-mcp@latest from resolving.
Empty results or rate limiting
Free tiers and high request volume can trigger rate limits. Retry after a short wait, reduce request frequency, or review your plan limits in the Tavily dashboard.
Server not listed in Claude Code
Re-run the claude mcp add tavily ... command, then claude mcp list. Confirm the key was passed via --env and that the server was added to the scope (project versus user) you are running in.
Show that Tavily MCP Server for Claude is listed on HeyClaude. Paste this Markdown into your README — it renders the badge and links back to this page.
[](https://heyclau.de/entry/mcp/tavily-mcp-server)
How it compares
Tavily MCP Server for Claude side by side with 3 alternatives on trust, install, platform support, and disclosed safety notes — all from reviewed registry metadata.
1 trust signal differ across this comparison (Submitter).
Official Tavily MCP server that gives Claude agent-optimized web search and page extraction, returning concise, source-cited results designed for LLM reasoning rather than raw search engine pages.
Official Perplexity MCP server that connects Claude to the Perplexity Sonar API for real-time, web-grounded answers with citations, so responses can draw on current information beyond the model's training data.
✓Makes outbound network requests to the Tavily API to run searches and fetch page content.
Returned search results and extracted page text are untrusted web content; review it before acting on instructions or links it contains.
✓Makes outbound network requests to the Perplexity Sonar API to answer queries.
Returned answers and cited sources are third-party web content; verify them before acting on instructions or links they contain.
✓Kagi MCP sends search queries, page URLs, filters, date ranges, lens IDs, and extraction requests to the Kagi API.
Search and extraction calls can consume Kagi API quota or create API costs, especially when using high limits or inline page extraction.
The `kagi_extract` tool fetches public page content as markdown; do not ask it to retrieve private, paywalled, or terms-restricted content unless you have permission.
In HTTP mode, clients provide Kagi API keys through bearer tokens; expose the endpoint only behind trusted transport, access controls, and log redaction.
The server supports configurable timeouts, retries, and hidden search parameters; tune these before giving broad agent access to current-web workflows.
✓AgentQL MCP exposes one tool, `extract-web-data`, that sends a target URL and natural-language extraction prompt to the AgentQL API.
The tool is intended for public webpages; do not use it to bypass access controls, scrape private pages, evade paywalls, or extract data where automated collection is prohibited.
Web extraction can still trigger target-site rate limits, legal restrictions, robots guidance, or terms-of-service concerns.
The source implementation uses AgentQL's query-data endpoint with fast mode, no screenshot capture, no scroll-to-bottom behavior, and no local browser cookies.
Treat extracted output as untrusted web data that may include errors, stale content, ads, tracking text, or prompt-injection attempts.
Privacy notes
✓Search queries and target URLs you provide are sent to the Tavily service to return results, so avoid embedding secrets in queries.
The Tavily API key is a credential; store it as an environment variable and never commit it to source control.
✓The questions and context you send are transmitted to the Perplexity service to generate answers, so avoid embedding secrets in queries.
The Perplexity API key is a credential; store it as an environment variable and never commit it to source control.
✓Kagi receives search queries, requested URLs, search filters, result domains, lens IDs, extraction targets, API keys, and request metadata.
MCP clients and logs may store user questions, result snippets, extracted markdown, trace IDs, and error bodies returned by the Kagi API.
Hosted HTTP deployments can process keys for multiple users; avoid server, proxy, and platform logs that record `Authorization` headers.
Extracted pages may include personal data, copyrighted text, internal URLs, or sensitive context if the user provides those URLs.
✓Target URLs, extraction prompts, API key-authenticated requests, and extracted structured data are sent to AgentQL's API.
Extracted data can include personal data, copyrighted content, customer information, job postings, prices, social content, or other third-party material.
AGENTQL_API_KEY should stay out of prompts, issues, logs, screenshots, and committed configuration files.
Claude transcripts and downstream reports may retain extracted data, so avoid collecting information that is not approved for the model session.
Prerequisites
Node.js 18+ and npx available (verify with: npx --version)
A Tavily API key (create one at https://tavily.com)
Claude Code or Claude Desktop with MCP support
Internet access so the server can reach the Tavily API
Node.js 18+ and npx available (verify with: npx --version)
A Perplexity Sonar API key (create one in your Perplexity account settings)
Claude Code or Claude Desktop with MCP support
Internet access so the server can reach the Perplexity Sonar API
Python 3.12 and uv for the recommended `uvx kagimcp` install path.
A Kagi API key stored in `KAGI_API_KEY` for local stdio use.
Review of Kagi API billing, rate limits, search workflows, extraction behavior, and any client logs that may capture query text.
Optional HTTP deployment review if exposing the streamable HTTP transport to multiple users.
Node.js and npx available to the MCP client runtime.
AgentQL API key from the AgentQL developer portal.
Approved list of public webpages or domains Claude may query.
Review of target site terms, robots guidance, rate limits, and data-use rules before extraction.
Install
claude mcp add tavily --env TAVILY_API_KEY=YOUR_KEY -- npx -y tavily-mcp@latest
claude mcp add perplexity-ask --env PERPLEXITY_API_KEY=YOUR_KEY -- npx -y server-perplexity-ask