> ## Documentation Index
> Fetch the complete documentation index at: https://www.truefoundry.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Set up Tavily MCP Server
> Connect Tavily, an AI-optimized web search API for RAG agents, as a fully managed MCP server on the TrueFoundry MCP Gateway.
[Tavily](https://docs.tavily.com) is an AI-optimized search API built for RAG agents and LLM applications, returning clean, structured results ready to pass straight into a prompt. TrueFoundry provisions it as a [TrueFoundry Managed MCP](/docs/ai-gateway/mcp/tfy-managed-mcp-server) - the platform manages the server URL, auth, scopes and credentials, so your agents can use Tavily without any configuration or infrastructure setup on your end.
## Using Tavily through the MCP Gateway
The [TrueFoundry MCP Gateway](/docs/ai-gateway/mcp/mcp-overview) centralizes access to Tavily - agents connect through one endpoint instead of maintaining separate server configurations per client. As a [TrueFoundry Managed MCP](/docs/ai-gateway/mcp/tfy-managed-mcp-server), Tavily's URL, outbound auth, scopes, and credentials are owned and kept current by the platform.
When you use Tavily MCP server through the Gateway, TrueFoundry provides:
* **Authentication** — Clients [authenticate inbound](/docs/ai-gateway/mcp/mcp-gateway-auth-security#inbound-authentication) to the Gateway with a Personal Access Token, Virtual Account, or IDE OAuth flow. For outbound access, the Gateway runs the [OAuth2 Authorization Code flow](/docs/ai-gateway/mcp/mcp-gateway-auth-security#outbound-authentication) so each user authorizes their own Tavily account; tokens are stored, refreshed and injected per user on tool calls.
* **Access control** — [Collaborators and role-based policies](/docs/ai-gateway/mcp/mcp-gateway-auth-security#access-control) define who can use the server and which tools they can invoke. Enable or disable individual Tavily tools from the server detail page.
* **Observability** — Tool calls are traced with caller identity, tool name, inputs, and latency. Monitor server- and tool-level usage in [MCP Metrics](/docs/ai-gateway/analytics-mcp-metrics) and export traces to your observability stack via OpenTelemetry.
* **Guardrails** — Apply [pre-tool and post-tool guardrails](/docs/ai-gateway/guardrails-overview) on MCP tool calls. Pre-tool checks run before execution and can block the call; post-tool checks run on the response before it returns to the agent.
## Adding Tavily MCP server to TrueFoundry
Navigate to **MCP Servers** in the TrueFoundry sidebar and click **Add new MCP Server**. On the next screen, select **Connect TrueFoundry Managed MCPs** - this opens the catalog of pre-vetted, fully configured MCP servers managed by TrueFoundry.
Find **tavily** in the catalogue and click **+ Add**. TrueFoundry provisions the server with no additional configuration and opens the Tavily MCP server detail page.
To authenticate, click **Connect Now**. A popup appears, review the details and click **Approve** to complete the authorization flow.
After approving, the **Tools** tab lists Tavily's available tools (search, extract, crawl, map, research, and more). Click **Try** on any tool to test it and inspect the JSON output before using it in an agent.
## Connecting to an MCP Client
Open the **How To Use** tab on the Tavily server detail page for your tenant-specific Gateway URL and ready-to-paste client snippets - don't build the endpoint manually.
The tab includes snippets for Claude Code, VS Code, Claude Web, Claude Desktop, Cursor, Windsurf, Codex, and the Python and TypeScript MCP SDKs. Use **Show API Key** if your client requires a Gateway token in a header.
## Using the Tool Playground
Before deploying an agent, you can test any Tavily MCP server tool directly in TrueFoundry:
On the Tavily MCP server detail page, click **Try** next to any tool - for example `tavily_search`.
For `tavily_search`, the key parameters are:
| Parameter | Description |
| --------------------------- | -------------------------------------------- |
| **Query** (required) | The search string |
| **Max results** | Number of results to return (default 5) |
| **Search depth** | `basic`, `advanced`, `fast`, or `ultra-fast` |
| **Topic** | `general` or a domain-specific filter |
| **Time range** | Restrict results to a recency window |
| **Include/Exclude domains** | Allowlist or blocklist specific sites |
Click **Execute Tool** and inspect the JSON output in the right panel. The output contains a `results` array where each item includes a `url`, `title`, `content` snippet, and a relevance `score`. The `tavily_research` tool returns a synthesized `answer` field in addition to the raw results.
## Tool Metrics
The **Tool Metrics** tab on the Tavily server detail page tracks how agents use each tool:
* **Invocation count** - which tools are called most often
* **Latency** - how long each tool takes to respond
* **Error rates** - which tools are failing in production
See [MCP Metrics](/docs/ai-gateway/analytics-mcp-metrics) for server- and tool-level dashboards across your Gateway.
## Disabling Individual Tools
On the **Tools** tab, toggle off any Tavily tools your agents don't need:
* **Disabled tools** are hidden from MCP clients and cannot be invoked.
* **Enabled tools** remain available to agents as usual.