Provides tools for interacting with Umami Analytics (both Cloud and self-hosted instances) to retrieve website lists, summary statistics, time-series pageview data, metric breakdowns (by URL, referrer, browser, etc.), and real-time active visitor counts.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Umami MCP ServerHow many active visitors are on my website right now?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Umami MCP Server
An MCP server for Umami Analytics — works with both Umami Cloud and self-hosted instances.
Zero dependencies. No cloning, no install steps — just point your MCP client at it.
Why?
Existing Umami MCP servers either don't support Umami Cloud (which uses API key auth, not username/password) or are broken and bloated (torch, faiss, sentence-transformers for… analytics?).
This server talks to the Umami API and exposes 5 tools over MCP. Pure Python, zero dependencies.
Tools
Tool | Description |
| List all tracked websites |
| Summary stats: pageviews, visitors, bounces, time on site |
| Time-series pageview data (bucketed by minute/hour/day/month/year) |
| Breakdown by URL, referrer, browser, OS, device, country, event, etc. |
| Number of currently active visitors (last 5 minutes) |
Quick Start
1. Get your credentials
Umami Cloud: Go to Settings → API Keys in your Umami Cloud dashboard and create an API key.
Self-hosted: Use the username and password you log in with.
2. Add to your MCP client
No cloning required — uvx fetches and runs it directly from GitHub.
Requires Install with:
curl -LsSf https://astral.sh/uv/install.sh | sh
Claude Desktop / Claude Code
Add to your MCP config (~/.claude.json, Claude Desktop settings, etc.):
Umami Cloud:
Self-hosted:
Cursor
Add to .cursor/mcp.json in your project root (or global settings):
VS Code (Copilot)
Add to your VS Code settings.json:
BoltAI
Go to Settings → MCP Servers → Add Server, then enter:
Command:
uvxArguments:
--from git+https://github.com/lukasschmit/umami-mcp umami-mcpEnvironment Variables:
UMAMI_URL=https://api.umami.isUMAMI_API_KEY=your_api_key_here
Environment Variables
Variable | Required | Description |
| Self-hosted: Yes, Cloud: Optional | Base URL — defaults to |
| Cloud | API key from Umami Cloud dashboard |
| Self-hosted | Login username |
| Self-hosted | Login password |
| Optional | Cloudflare Access service token client ID (for protected self-hosted APIs) |
| Optional | Cloudflare Access service token secret |
| Optional | Custom User-Agent for outbound requests (default: |
Set either UMAMI_API_KEY (Cloud) or both UMAMI_USERNAME + UMAMI_PASSWORD (self-hosted). The server auto-detects which mode to use.
For convenience, UMAMI_URL may include /v1 (Cloud) or /api (self-hosted); suffixes are normalized automatically.
If your self-hosted Umami is behind Cloudflare Access, set both UMAMI_CF_ACCESS_CLIENT_ID and UMAMI_CF_ACCESS_CLIENT_SECRET so machine-to-machine MCP calls can pass Access checks.
get_metrics accepts both type="path" and type="url" for compatibility across Umami versions.
Usage Examples
Once connected, you can ask your AI assistant things like:
"What are my top pages this week?"
"Show me visitor trends for the last 30 days"
"Which countries are my visitors from?"
"How many people are on my site right now?"
"Compare this month's traffic to last month"
The assistant will call the appropriate tools with the right parameters.
How It Works
The server implements the Model Context Protocol over stdio (JSON-RPC, one JSON object per line). When an MCP client starts it, the server:
Reads JSON-RPC messages from stdin
Handles
initialize,tools/list, andtools/callmethodsMakes authenticated HTTP requests to the Umami API
Returns results as JSON text content
No background processes, no polling, no state beyond the auth token.
License
MIT