plausible-mcp
Enables teams to query Plausible Analytics from Claude using Google sign-in, with org-wide access managed through allowlisted domains or emails and no per-person API keys.
README
plausible-mcp
An MCP server that lets everyone in an organisation query Plausible Analytics from Claude.
One deployment serves the whole org. It holds a single Plausible API key server-side
and authenticates each person with Google, so nobody needs their own key and nobody
has to be granted access individually: put your Google Workspace domain in
ALLOWED_DOMAINS and anyone with an address there can connect the first time they
try. Colleagues outside that domain go in ALLOWED_EMAILS, one address at a time.
If you run analytics for a team and want them reading the numbers themselves rather than asking you for them, that is what this is for.
How authentication works
Claude registers itself with this server using OAuth 2.1 dynamic client registration. Google does not support dynamic client registration, so this server acts as its own authorization server and delegates the human login upstream to Google.
Google establishes who the person is. It does not decide whether they may read your
analytics. A middleware checks each caller's verified email address against
ALLOWED_DOMAINS and ALLOWED_EMAILS on every tool call. If you set neither, the
server refuses to start.
ALLOWED_DOMAINS is the part that makes this org-wide: it matches the domain of the
verified address exactly, so example.com admits alice@example.com but not
alice@evil-example.com or alice@example.com.evil.com.
To revoke someone, take them out of ALLOWED_EMAILS or ALLOWED_DOMAINS and
redeploy. Suspending their Google account stops them signing in again, but the
allowlist reads the claims from a token this server signed, so one already issued
keeps working until it expires. Rotating JWT_SIGNING_KEY invalidates every token
at once, at the cost of signing everybody out.
Tools
queryruns a Stats API v2 query. Grouping by different dimensions gives you timeseries, breakdowns and goal conversions from the same tool. To compare two periods, call it once per period.list_sitesreturns the site domains you can pass assite_id.usage_statsreports who has used the server, how often, and which accounts were refused.
Usage tracking
The allowlist middleware sees every caller's email address, because it has to check
it. It records a per-person call count, a first-seen and last-seen time, and a
per-tool breakdown to the volume at USAGE_PATH. Refused accounts are counted
separately, which is how you spot someone outside the allowlist trying to connect.
Read it with the usage_stats tool. Anyone on the allowlist can see everyone's
rows. Unset USAGE_PATH to turn tracking off.
Configuration
| Variable | Required | Purpose |
|---|---|---|
GOOGLE_CLIENT_ID / GOOGLE_CLIENT_SECRET |
yes | Google OAuth client used to sign people in |
PLAUSIBLE_API_KEY |
yes | Plausible key with Stats API access; shared by everyone |
BASE_URL |
yes | Public URL of this server; the Google redirect URI is BASE_URL + /auth/callback |
ALLOWED_DOMAINS / ALLOWED_EMAILS |
one of them | Who may query. Neither set means the server refuses to start |
JWT_SIGNING_KEY |
recommended | Signs this server's access tokens. Without it a restart signs everyone out |
CLIENT_STORAGE_PATH |
no | Where to persist OAuth client registrations. Unset means in-memory |
USAGE_PATH |
no | Where to record per-person usage. Unset turns tracking off |
See .env.example for the same list in copy-pasteable form.
Deploy
The repo ships a fly.toml, so the instructions below are for Fly.io,
but nothing in the server is Fly-specific: it is one container that wants a writable
directory and the environment above. Replace YOUR-APP throughout.
-
Create an OAuth client in the Google Cloud console. Choose Web application and set the authorized redirect URI to
https://YOUR-APP.fly.dev/auth/callback. The URI must match exactly. -
Create the app and its volume:
fly apps create YOUR-APP --org YOUR-ORG fly volumes create plausible_mcp_data --app YOUR-APP --region fra --size 1 --yesUse
fly apps createrather thanfly launch, which regeneratesfly.tomland would discard the volume mount and health check. -
Point
appinfly.tomlat your app name. -
Set the configuration.
BASE_URLandALLOWED_DOMAINSgo here rather than infly.tomlbecause they differ per deployment, and a placeholder committed to the repo would overwrite the real value on the next deploy:fly secrets set --app YOUR-APP \ BASE_URL=https://YOUR-APP.fly.dev \ ALLOWED_DOMAINS=your-domain.org \ GOOGLE_CLIENT_ID=xxx.apps.googleusercontent.com \ GOOGLE_CLIENT_SECRET=GOCSPX-xxx \ PLAUSIBLE_API_KEY=xxx \ JWT_SIGNING_KEY="$(openssl rand -hex 32)"Keep
JWT_SIGNING_KEYstable: changing it signs everyone out. -
fly deploy --ha=falseWithout
--ha=falseFly may start two machines. Each would get its own volume, so a sign-in that registers on one machine fails when the next request lands on the other.
Add it to Claude
To add it for a whole team, go to Settings > Connectors in claude.ai, choose
Add custom connector, and enter https://YOUR-APP.fly.dev/mcp. Each person
signs in with Google the first time they use it.
To add it to Claude Code:
claude mcp add --transport http plausible https://YOUR-APP.fly.dev/mcp
Run it locally
uv venv && uv pip install -r pyproject.toml
cp .env.example .env # then fill it in
set -a && source .env && set +a
python server.py
Add http://localhost:8000/auth/callback to the Google client's redirect URIs to test
the sign-in flow locally.
Run the tests with python test_server.py. To exercise the OAuth flow and the tools by
hand, use npx @modelcontextprotocol/inspector.
Licence
MIT — see LICENSE.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.