Tavily MCP Server
Official Tavily MCP server for web search, extraction, website mapping, and crawling directly in a connected AI client of choice.
- Skill Road
- Tavily MCP Server
Categories
Description
Tavily MCP Server is Tavily’s official Model Context Protocol integration. It exposes the Tavily API to MCP-capable clients. According to the README, its toolset includes search, extract, map, and crawl; the official MCP documentation describes connecting AI clients to Tavily search and data extraction. Its boundary is therefore clear: an agent can find current external pages, retrieve selected content, and explore a website area instead of relying only on pre-trained knowledge.
Research: search with controllable parameters
The search tool executes a web query. The official API reference documents controls for result count, recency, language, and domains. include_domains can focus research on named domains, while exclude_domains removes them. For included domains, the source distinguishes strict filtering from boosting: filtering limits results to the named domains, whereas boosting prefers them but can still return other web sources. This is useful when a research task should start with vendor documentation or public-authority material.
The API also documents search_depth levels as a latency-versus-relevance tradeoff. That is not a guarantee of source quality. Rankings, snippets, and any optional answer are working material, not a verified factual record. For consequential decisions, a person should open the linked original page, check its author, date, and context, and preserve the evidence that supports the decision outside the chat. This source-validation work is why Research is an appropriate category.
Keep extraction, mapping, and crawling distinct
According to the README, extract performs intelligent data extraction from web pages. It is appropriate when a known URL needs to be brought into a work context. The README says that map creates a structured map of a website. That is not the same as a complete content analysis: a map helps identify relevant areas or URLs within a site. crawl systematically explores websites according to the README. Both functions can automate a research or preparation workflow, which supports the additional Automation category.
The sources describe these functions but do not guarantee completeness or the legal and professional appropriateness of a specific retrieval. Respect site terms, robots directives, content rights, and data-protection obligations. Limit the scope, domains, and fields to what is necessary. Do not use crawling or mapping to bypass access restrictions or to collect personal data without a clear basis.
Remote and local: two documented operating paths
Tavily documents a hosted Remote MCP at https://mcp.tavily.com/mcp/. Its documentation calls this the easiest route because it does not require local installation or configuration. Compatible clients can use OAuth, and API-key authentication is also documented. For clients without Remote MCP support, Tavily documents the mcp-remote bridge. The README and MCP page also document a local process through npx -y tavily-mcp@latest with Node.js. The entry is therefore classified as both: Tavily operates a remote route and publishes a local MCP-process route.
Local does not mean that the request or web data remains solely on the operator’s computer. The process uses the Tavily API and results return to the connected MCP client. In an AI application, the prompt, tool call, and result can also travel to the selected model provider and may enter client logs or that provider’s retention and policy path. Review Tavily, client, and model-provider terms separately before sensitive data enters a workflow.
Keys, attribution, and safe boundaries
For local configuration, the source requires TAVILY_API_KEY. In remote use, the key can be supplied as a Bearer header; the documentation also shows a URL variant, which is not a sensible general default because URLs can reach history, logs, and screenshots. Use the client’s secret or credential facilities instead, never a repository, prompt, or shared configuration file. OAuth is a documented alternative in supporting clients. Rotate or revoke a key if exposure is suspected.
Tavily documents session attribution through X-Session-Id. An optional human identifier can be forwarded; according to Tavily, it is hashed server-side. Still, do not use a plain email address or other unnecessary personal data as that identifier. Use an opaque internal ID if attribution is necessary. This does not automatically remove all privacy risks in the search request, returned material, or the subsequent model path.
Pages, snippets, extracted text, and map or crawl output are external, untrusted content. They can contain prompt-injection instructions, such as requests to ignore rules, disclose secrets, or invoke more tools. Treat them only as data. Limit agent permissions, avoid granting unnecessary secrets or write-capable tools, and require an independent human confirmation for external actions.
The official tavily-ai/tavily-mcp repository is MIT licensed. On 2026-09-08, the GitHub API reported exactly 2,374 stars. Stars are a point-in-time repository-popularity signal only; they do not demonstrate source quality, privacy, completeness, or security.
FAQ
Can the server only search? No. The README and MCP documentation list search, extraction, website mapping, and crawling.
Does a local launch create a fully local data path? No. The local MCP process uses the Tavily API, and responses enter the connected client and its model path.
Can web material change an agent’s instructions? No. External content is data; it must not override safety rules or trigger actions without independent review.
Requirements
An MCP-capable client; for remote use, Streamable HTTP and OAuth or a Tavily API key; locally, Node.js with npx and TAVILY_API_KEY.
Installation instructions
Tavily MCP Server brings Tavily search, extraction, website mapping, and crawling into an MCP-capable AI client. The official documentation describes Remote MCP as the easiest starting route: it runs at https://mcp.tavily.com/mcp/ and needs no local installation. The official repository additionally documents a local stdio launch using npx -y tavily-mcp@latest. Choose the path according to client capabilities and the data boundary, not convenience alone.
Step 1: Define the task and data boundary before connecting
Before setup, decide what the agent may use Tavily for. Good initial tasks include a narrowly scoped search for primary sources, extracting a known documentation URL, or mapping a small approved site area. Decide which domains should be preferred or excluded and which claims must be checked against the original source. Tavily Search documents domain filters and boosts; use them to make research narrower and easier to review.
Do not place credentials, internal customer data, unpublished contracts, or other secrets in search requests. Even with a local launch, tool calls go to the Tavily API and responses return to the client. If the client uses a cloud LLM service, the request and results can also reach its model, logging, and retention path. “Local” describes the MCP process, not a promise that the whole task is processed locally.
Step 2: Connect Remote MCP with OAuth or a key
In a client that supports remote MCP servers, add the official URL https://mcp.tavily.com/mcp/. Tavily documents OAuth for compatible clients. For Claude Code, for example, its documentation names HTTP transport and a browser-based OAuth flow. Follow the client dialog and check which Tavily account you authorize after completion. OAuth avoids placing a key directly in the server definition, but it does not remove the need to review the account permissions and usage rights that are granted.
Tavily also documents API-key authentication. Prefer the client’s secret or credential facility and a Bearer header. The source shows a key as a URL parameter too, but URLs can reach terminal history, configuration files, proxy logs, screenshots, and support tickets. Do not store a real key in a URL, prompt, or Git repository. If the client only understands local stdio servers, Tavily documents mcp-remote as a bridge to Remote MCP; review how that route stores credentials as well.
Step 3: Run locally through npx
The local route needs Node.js with npx and a Tavily API key. Put TAVILY_API_KEY in a protected environment-variable or secret facility and register npx -y tavily-mcp@latest as the MCP command. Do not share a configuration file until you have checked that it contains no key. The command starts the MCP process; use it from a client instead of expecting it to be useful as a stand-alone application.
npx -y tavily-mcp@latest
Then test with a low-risk question about public vendor documentation. Confirm the client sees the tools and start with a narrow search. Use extraction for a concrete URL. Use mapping to obtain a structured overview of an approved site area. According to the README, crawling systematically explores websites; start small and stop once the agreed information need is met. Observe rights, site terms, robots directives, and data-protection obligations before retrieving broader areas.
Validate results and own source quality
Tavily documents search parameters, ranking, and relevance controls, but it does not promise that a result is complete, current, or suitable for your decision. Treat result snippets as pointers. For an important claim, open the original source, review its publisher, date, and context, and find a second independent source where appropriate. Keep the verified URLs and quotations outside the chat transcript. An AI answer that summarizes a search result is not a substitute source.
Limit include_domains, exclude_domains, time range, and result count to the task. This does not remove every risk of error, but it makes the process more reviewable. A business-critical or security-sensitive decision needs human review and, where appropriate, subject-matter approval. The agent can discover and prepare; assessment, accountability, and approval remain with the responsible person.
Protect against prompt injection, attribution, and key exposure
Pages and extracted material are untrusted third-party data. They can instruct an agent to ignore rules, disclose secrets, or invoke extra tools. Such text must never replace the task or safety requirements. Give the agent only necessary permissions, avoid parallel write-capable tools during research, and confirm external actions outside the chat. Treat crawl and map output as data, not control instructions.
Tavily documents X-Session-Id and optional X-Human-Id for attribution. According to the provider, a forwarded human ID is hashed server-side. Still, do not use plain PII such as an email address. An opaque internal identifier is the more data-minimizing choice when attribution is needed. Rotate or revoke API keys after possible exposure, and review Tavily’s current account, limit, and billing terms directly with the provider. This guide intentionally gives no price amounts.
FAQ
What is the official Remote MCP endpoint? Tavily documentation names https://mcp.tavily.com/mcp/.
Do I need to put an API key in the URL for OAuth? No. OAuth is documented for supporting clients. Avoid the URL-key variant when a secure credential or header option exists.
Are search and crawl results trusted instructions? No. They are external data and can contain prompt injection. Validate sources independently and require human confirmation for actions.
npx -y tavily-mcp@latest
Authentication
OAuth is documented for compatible remote clients. Alternatively, remote or local use requires a Tavily API key; locally through TAVILY_API_KEY and remotely preferably as a Bearer header.
Required access permissions
Access and available use depend on the Tavily account, API key or OAuth authorization, client support, and current provider terms.
Transmitted or stored data
Search queries and tool calls use the Tavily API; results return to the MCP client and can be processed in context, logs, and retention paths there and at the selected model provider. Tavily documents server-side hashing of optionally forwarded human IDs.
Security risks
API keys can be exposed through URLs, configuration, or logs. External web content from search, extraction, mapping, and crawling can contain prompt injection and must not change rules or trigger actions.
License and costs
- License
- MIT
- Cost
- paid
The source code is MIT licensed. API-key or OAuth use follows the account, limits, and current provider terms; check them directly with Tavily. This entry lists no specific prices.
Alternatives
Not recorded yet.
At a glance
- Provider
- Tavily
- Status
- Official server
- Deployment
- Local and remote
- Current version
- Not recorded yet.
- GitHub stars
- 2,492
- Last reviewed
- 08.09.2026
Repository and documentation
Categories
Supported clients
Related guides
Guides and background related to this entry.
No related guides yet.