> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tiro.ooo/llms.txt
> Use this file to discover all available pages before exploring further.

# Install Tiro MCP in your agent

> One prompt — your agent installs the Tiro MCP server, signs you in through OAuth, and verifies the connection.

Paste the block below into Claude Desktop, Claude Code, Cursor, ChatGPT, or any MCP-compatible client. The agent reads the instructions, connects to `https://mcp.tiro.ooo/mcp`, and signs you in through your browser on first use.

**Compatible clients**  ·  Claude Desktop  ·  Claude Code  ·  Cursor  ·  ChatGPT  ·  VS Code (Copilot)  ·  any streamable-HTTP MCP client

## Copy this prompt

```text The Tiro MCP install prompt theme={"system"}
Install the Tiro MCP server in this workspace.

  Server URL :  https://mcp.tiro.ooo/mcp
  Transport  :  HTTP (streamable)
  Auth       :  OAuth (preferred). On first connection, open a browser
                for Google sign-in. Tokens cache locally for 180 days.

If this client cannot run a browser flow, ask me for a Tiro API key
from https://platform.tiro.ooo/dashboard/api-keys and attach the
header:

  Authorization: Bearer <TIRO_API_KEY>

For stdio-only clients, wrap the endpoint with mcp-remote:

  npx -y mcp-remote https://mcp.tiro.ooo/mcp

After connecting, list the Tiro tools that loaded and call
`auth_status` to confirm the session is live. If a step fails, show
me the exact error and the step where it happened — do not retry
silently.
```

<Tip>
  The prompt is intentionally **client-agnostic**. Your agent picks the right config file (`claude_desktop_config.json`, `.vscode/mcp.json`, Cursor settings, etc.) based on where it's running.
</Tip>

## What your agent will do

1. Detect the active MCP client and open its config.
2. Add a `tiro` server entry pointing at `https://mcp.tiro.ooo/mcp`.
3. Restart the client (or prompt you to) and complete OAuth in your browser.
4. Call `auth_status` to verify the connection and report which tools are available.

If anything fails, the agent surfaces the exact error so you know which step to retry — no silent loops.

## Privacy & token handling

* OAuth tokens are cached in your local MCP client config only — never sent anywhere else.
* API keys (if used as fallback) are stored in plain text inside that same config file; keep it out of shared dotfiles.
* Revoke either at any time from the [API Keys dashboard](https://platform.tiro.ooo/dashboard/api-keys) or by signing out from your client.

***

Prefer to wire it up yourself?  → [Manual setup per client](/en/developers/mcp/setup)   ·   [Available tools](/en/developers/mcp/tools/overview)   ·   [Troubleshooting](/en/developers/mcp/troubleshooting)