Arcane MCP is a Model Context Protocol server that gives AI agents full control over Docker — containers, images, volumes, networks, Compose projects, and more.
AI agents talk to Docker through a lightweight MCP server. The Arcane API handles the infrastructure — Arcane MCP handles the protocol.
13 module categories covering the full Docker API surface — from containers to vulnerability scanning.
Create, start, stop, kill, pause, exec, logs, stats, auto-update — full container lifecycle.
Pull, push, build, tag, prune, scan vulnerabilities, export, view history and attestations.
Create, inspect, backup, restore, browse files, read/write inside volumes, prune unused.
Create, connect, disconnect, inspect topology, prune unused networks across environments.
Deploy, update, scale, restart, archive, inspect runtime state — full Compose lifecycle.
Scan images, list CVEs, ignore/unignore findings, check scanner health across the fleet.
Docker info, version, prune, health check, upgrade management, bulk container ops.
Manage container registries — create, update, test connectivity, delete.
Trigger, list, create, update, delete inbound webhooks for event-driven automation.
List and cancel background activities, browse system events, delete stale entries.
Manage multiple Arcane environments — local Docker, remote agents, multi-tenant setups.
Run, monitor, and review the container updater — keep images fresh across the stack.
List all port mappings across every container and environment in one call.
Install, configure, connect. Three steps and your agent talks to Docker.
Not on PyPI. Install directly from GitHub.
Set your Arcane instance URL and API key. Drop them in a .env or set as env vars.
Add to your opencode.jsonc and restart — all 123 tools appear automatically.
Docker operations are powerful. Every destructive action requires explicit confirmation.
Every tool that removes, prunes, kills, restores, or overwrites requires
confirm=True — without it, the tool returns a warning.