RudderStack MCP
Serveur MCP hébergé qui permet aux assistants IA de déboguer les pipelines RudderStack : erreurs de livraison, chargements d'entrepôt, violations du plan de suivi et transformations.
Serveur MCP hébergé
npx add-mcp 'https://mcp.rudderstack.com/mcp'S’installe dans Claude Code, Codex, Cursor et plus
Documentation
This guide covers the steps to connect your AI assistant to RudderStack MCP for pipeline debugging and management.
Prerequisites
- You must have an active RudderStack account
- MCP-compatible client (Claude, Codex, Cursor, VS Code with GitHub Copilot, etc.)
Client setup
RudderStack MCP is a hosted remote server that uses Streamable HTTP at https://mcp.rudderstack.com/mcp. The server uses browser-based OAuth, so don’t add a RudderStack access token, client secret, or other static credential to your MCP client configuration.
All configurations below are global (available across all projects) unless noted otherwise. Refer to your MCP client’s documentation to change the scope.
Other MCP clients
For any client that supports remote MCP servers:
- Add a remote server named
rudderstack. - Select Streamable HTTP as the transport, if the client asks for one.
- Enter
https://mcp.rudderstack.com/mcpas the server URL. - Connect to the server and complete the OAuth flow in your browser.
Configuration field names vary by client. You only need the hosted server URL — don’t configure a local command or static credentials.
Claude.ai
- Navigate to claude.ai/settings > Connectors > Add custom connector.
- Enter the URL:
https://mcp.rudderstack.com/mcp
Claude Code
Run the following command:
claude mcp add --transport http rudderstack https://mcp.rudderstack.com/mcp
To make this available across all projects, add --scope user:
claude mcp add --transport http --scope user rudderstack https://mcp.rudderstack.com/mcp
Authenticate by running /mcp in an active Claude Code session and following the OAuth flow.
Claude Desktop
- Navigate to Settings > Connectors > Add custom connector.
- Enter the URL:
https://mcp.rudderstack.com/mcp
Codex
- Add the following to your
~/.codex/config.tomlfile:
[mcp_servers.rudderstack]
url = "https://mcp.rudderstack.com/mcp"
- Authenticate using the following command — this takes you through the OAuth flow.
codex mcp login rudderstack
Cursor
- Open Cursor Settings > MCP > New MCP Server.
- Add the following to
~/.cursor/mcp.json:
{
"mcpServers": {
"rudderstack": {
"url": "https://mcp.rudderstack.com/mcp"
}
}
}
Visual Studio Code (GitHub Copilot)
Add the following to .vscode/mcp.json in your project root. Note that this is project-scoped — you will need to add it to each project where you want to use RudderStack MCP.
{
"servers": {
"rudderstack": {
"url": "https://mcp.rudderstack.com/mcp"
}
}
}
Authentication
On first connection, a browser window opens for RudderStack OAuth login. Authorize access to your workspace and your client automatically receives access.
Tokens refresh automatically without requiring manual re-authentication.
Verify your connection
Ask your AI assistant:
“Give me an overview of my RudderStack workspace.”
You should see your sources, destinations, and connections listed.
Troubleshooting
| Issue | Solution |
|---|---|
| OAuth login window doesn’t open | Make sure you have completed the OAuth authorization flow for your client |
| Tools not showing up | |
| “Workspace not found” error | Ask “Show me my available workspaces” and switch to the correct workspace |
| Connection drops or session errors | Restart the MCP client to establish a fresh session |