Mcparr
FreeNot checkedA self-hosted MCP gateway that lets AI assistants drive your media services like Radarr and Sonarr through a single authenticated endpoint.
About
A self-hosted MCP gateway that lets AI assistants drive your media services like Radarr and Sonarr through a single authenticated endpoint.
README
MCPify your *arr stack. mcparr is a self-hosted Model Context Protocol gateway that lets AI assistants (Claude, ChatGPT, GitHub Copilot, Cursor, ...) drive your media services - Radarr, Sonarr, and more - through a single, authenticated MCP endpoint.
What it does
mcparr runs as one Docker container exposing two ports:
- 7475 - config UI: a small web app to add service connections, test them, toggle tools, and copy the MCP connection details. Protected by an admin password.
- 7474 - MCP endpoint: a Streamable HTTP MCP server that AI clients connect to. Always requires a bearer token.
Each enabled service contributes a set of namespaced tools (for example
radarr_search_movies, sonarr_list_series). Toggling a service in the UI adds
or removes its tools live, without restarting the container.
Status
Early development. The v1 surface ships the gateway core plus Radarr and Sonarr; more services (Prowlarr, Plex, Overseerr/Seerr, Maintainerr, Lidarr) are planned.
Quick start
docker compose up -d
Then:
- Open
http://localhost:7475and set an admin password. - Add a service (base URL + API key), test the connection, and enable it.
- Open the Connect page to copy the MCP URL and bearer token into your AI client.
The config UI is published on localhost only by default. The MCP endpoint is exposed but always requires the token.
Connecting a client (VS Code / GitHub Copilot)
{
"servers": {
"mcparr": {
"type": "http",
"url": "http://your-host:7474/mcp",
"headers": { "Authorization": "Bearer <token-from-connect-page>" }
}
}
}
For Claude Desktop, use the mcp-remote bridge (snippet on the Connect page).
Hosted connectors (Claude.ai, ChatGPT) require HTTPS - put mcparr behind a TLS
reverse proxy.
Release channels
The image is published to GHCR under three rolling tags. Set image: in
docker-compose.yml to the channel you want:
| Tag | Tracks | Use it when |
|---|---|---|
:latest |
Newest stable release | Default. You want tested, stable builds. |
:beta |
Newest pre-release | You want to preview fixes before :latest. |
:dev |
Every merge to main (bleeding edge) |
You are testing unreleased work; may break. |
For reproducible deployments, pin an exact version instead - for example
ghcr.io/phillyurbs/mcparr:0.1.1. Stable releases also publish the moving
:0, :0.1 major/minor tags.
Configuration
All state lives in the /data volume (SQLite database, encryption key, MCP
token, audit log). Useful environment variables:
| Variable | Default | Purpose |
|---|---|---|
MCPARR_HOST |
127.0.0.1 |
Bind address for both ports |
MCPARR_UI_PORT |
7475 |
Config UI port |
MCPARR_MCP_PORT |
7474 |
MCP endpoint port |
MCPARR_DATA_DIR |
/data |
State directory |
MCPARR_SECRET_KEY |
(generated) | Fernet key for secrets at rest |
MCPARR_MCP_TOKEN |
(generated) | MCP bearer token (env wins; rotation disabled) |
MCPARR_ADMIN_PASSWORD |
(unset) | Set the admin password headlessly on first run |
MCPARR_SEED_FILE |
<data>/seed.yaml |
Optional service seed file |
You can pre-populate services declaratively with a seed file - see config.yaml.example.
Security
The MCP endpoint always requires a token; the UI always requires a password. Service keys are encrypted at rest. See SECURITY.md and the threat model for details.
Development
python -m venv .venv && source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -e ".[dev]"
pytest -q
ruff check .
python -m mcparr.main # runs both servers locally
Translations are managed with Babel (babel.cfg); compile catalogs with
pybabel compile -d mcparr/ui/locales -D messages.
Branch workflow
Work on a short-lived branch and open a pull request into main; CI runs on the
PR. No :dev image is built from branch pushes, so committing early and often is
harmless. Merging the PR into main produces one :dev build (squash-merge
keeps history tidy). Add [skip ci] to a commit message to skip workflows for
that commit, and docs-only changes (Markdown, docs/, LICENSE) never trigger
an image build.
Cutting a release
Releases are version-free for you: the Release workflow computes the next version, tags the commit, and builds the channel image.
- Open the Actions tab (or the GitHub Actions panel in VS Code) and run the Release workflow.
- Pick a channel (
betaorstable) and a bump (autoby default). - The workflow tags the commit and publishes the image.
With bump = auto:
- beta patch-bumps the latest stable and starts/continues a
-beta.Nseries -v0.1.0->v0.1.1-beta.1->v0.1.1-beta.2-> ... - stable promotes an in-flight beta (
v0.1.1-beta.2->v0.1.1), or patch- bumps if no beta is pending.
Use bump = minor or major to override the base when a release is bigger than
a patch. The version math lives in scripts/next_version.py
and is covered by tests/test_next_version.py.
Adding a service
Create a ServiceModule subclass under mcparr/services/, declare its metadata
and tools, and decorate it with @register. The base class provides the HTTP
client, pagination, result shaping, and error mapping. Radarr and Sonarr are the
reference implementations.
License
See LICENSE.
Installing Mcparr
This server has no published package — it is built from source. Open the repository and follow its README.
▸ github.com/PhillyUrbs/mcparrFAQ
Is Mcparr MCP free?
Yes, Mcparr MCP is free — one-click install via Unyly at no cost.
Does Mcparr need an API key?
No, Mcparr runs without API keys or environment variables.
Is Mcparr hosted or self-hosted?
A hosted option is available: Unyly runs the server in the cloud, no local setup required.
How do I install Mcparr in Claude Desktop, Claude Code or Cursor?
Open Mcparr on unyly.org, pick your client tab (Claude Desktop, Claude Code, Cursor) and press Install — the config is generated automatically, no JSON editing.
Related MCPs
Omni Video
An MCP server that transforms LLM-enabled IDEs into professional video editors by pre-processing footage into text proxies, generating motion graphics via HTML/
by buildwithtazaARA
Generate images, video and audio from any AI agent — one connector.
by ARAYouTube
Transcripts, channel stats, search
by YouTubeEverArt
AI image generation using various models.
by modelcontextprotocolCompare Mcparr with
Not sure what to pick?
Find your stack in 60 seconds
Author?
Embed badge for your README
Browse similar
All media MCPs
