Umami

You can connect Umami Analytics to any MCP client using the Umami MCP Server. It enables you to run the server locally or remotely and access Umami data through various MCP clients like Claude Desktop, VS Code, Cursor, Windsurf, Zed, and more, creating a seamless bridge between Umami and your preferred tooling.

How to use

Set up the MCP server in either local or remote mode, then configure your MCP client to connect using the appropriate transport. You can run the server locally with your Umami instance credentials or connect through a hosted MCP endpoint that forwards requests to your Umami instance. Once connected, you can issue prompts or commands in your MCP client to retrieve analytics, traffic data, user insights, real-time monitoring, and content analytics from Umami.

How to install

Prerequisites you need before installation: a computer with a compatible operating system and a Go toolchain installed, or Docker if you prefer containerized usage. You can also use a prebuilt binary for quick setup.

go install github.com/Macawls/umami-mcp-server@latest
# Or a specific version
# go install github.com/Macawls/umami-mcp-server@v1.0.3

Configure your MCP client to connect

You can connect from both local and remote MCP clients. For a local setup, you run the server binary and point your client to the local process. For a remote setup, you can use a hosted MCP endpoint that forwards to your Umami instance.

Additional configuration and usage notes

Transport modes control how the MCP server communicates. The default is stdio for local clients, while HTTP is available for self-hosting or remote deployments. You can choose the transport by setting environment variables or by using explicit configuration from your MCP client.

To run in HTTP mode, you start the server with the HTTP transport and specify the Umami instance and credentials in the request, allowing any MCP client that supports Streamable HTTP to connect.

Troubleshooting

If the server won’t run, verify you have a compatible Go toolchain or Docker installed and ensure the Umami instance URL is accessible. Check credentials for correctness and confirm that the MCP client can reach the server. If tools do not appear in the MCP client, verify the binary path and run the server directly to observe errors.

Security and best practices

Keep credentials secure. Prefer using secure prompts in clients that support secrets and avoid exposing usernames or passwords in plain text configurations. When using HTTP transport, ensure the connection is over HTTPS to protect sensitive data in transit.

Tools and endpoints

The following tools are exposed by the MCP server to retrieve and analyze Umami data: get_websites, get_stats, get_pageviews, get_metrics, get_active. Use these tools through your MCP client to list websites, view visitor statistics, analyze page traffic over time, explore browsers and devices, and monitor current active visitors.

Supported MCP connection options

HTTP (remote) connection: You can connect to a hosted MCP endpoint that forwards requests to your Umami instance. The remote URL is used directly in your MCP client configuration.

STDIO (local) connection: Run the MCP server binary locally and connect via standard input/output in your MCP client. This is the default mode for many clients.

Transport modes in detail

Stdio is the default transport used by local MCP clients such as Claude Desktop, VS Code, Cursor, Windsurf, Zed, and Cline. HTTP exposes a /mcp endpoint that speaks Streamable HTTP for remote usage.

Available tools

get_websites

List all your websites configured in Umami.

get_stats

Retrieve visitor statistics such as sessions, pageviews, and visitors.

get_pageviews

View site traffic over time by pageviews and dates.

get_metrics

See browsers, countries, devices, and other demographic metrics.

get_active

Show current active visitors on your sites.