MCP server · Docker & Podman

Talk to your
media server stack

Natural language management for Plex, Sonarr, Radarr, SABnzbd — via Claude. No dashboards, no SSH.

Get started Tools reference →
Claude · arr-mcp
you    Which containers are using the most memory?
arr-mcp Running container_stats()…
Top by memory:
plex         1.8 GB  (of 4 GB limit)
sabnzbd      312 MB
sonarr       188 MB
you    Restart sabnzbd and tail its logs
arr-mcp container_restart(“sabnzbd”) → ok · fetching logs…
[2026-06-02 14:32:01] SABnzbd 4.3.2 starting…
[2026-06-02 14:32:03] Web interface on port 8080
What it does

Everything your stack needs, in plain English

arr-mcp exposes your container runtime and media server config as MCP tools, so Claude can act on your behalf.

Container lifecycle

List, start, stop, restart and remove containers. Check CPU, memory and network stats across your whole stack.

Compose files

Read and write compose.yaml for any stack. Claude can propose changes and apply them in one turn.

Filesystem ops

Scoped to allowed roots. List directories, read and write config files, check disk usage — all sandboxed.

Log tailing & search

Tail the last N lines from any log file, or search them case-insensitively. Diagnose issues without SSH.

Docker & Podman

Runs against Docker Engine, Docker Compose, or rootless Podman with Quadlets — including systemd.

Connects to Claude

Works with Claude.ai remote MCP or Claude Desktop via mcpproxy. Add it to your integrations in minutes.

Tools reference

Available tools

All tools available today — more coming with the host-side helper agent.

container_list()
All containers with status, uptime, and ports
container_start(name)
Start a stopped container
container_stop(name)
Stop a running container
container_restart(name)
Restart a container
container_remove(name, confirm)
Remove a container — requires confirm=True
container_logs(name, lines)
Fetch last N log lines from a container
container_stats()
CPU, memory, and network stats for all running containers
Getting started

Up in three steps

Supports Docker Engine, Docker Compose, and rootless Podman with Quadlets.

✓Docker Engine
✓Docker Compose
✓Podman + Quadlets
1

Run the container

Pull and start arr-mcp with your API key and container socket mounted. Exposes port 8081.

ghcr.io/ryanbrinn/arr-mcp:latest
2

Verify the health endpoint

Confirm the server is running and reachable before connecting Claude.

curl http://localhost:8081/health
3

Connect Claude

On Claude.ai go to Settings → Integrations and add your server URL with a Bearer token. For Claude Desktop, bridge via mcpproxy.

http://your-server-ip:8081/mcp