v1.0.6MCPread-onlydocker

🐳 dockroot-mcp

Virtual Docker Root for AI Agents. Give your agent read-only power over its own Docker stack — without host root.

docker_ps · docker_logs · docker_inspect · docker_stats · docker_networks
plus the full tavern connectivity toolkit (status, self_check, wire, rediscover, models, chat).

Why this exists

AI agents get asked "why is my Open WebUI broken?" constantly. Without docker access they flail — they invent commands, read the wrong configs, or hallucinate fixes. With dockroot-mcp the agent can:

List the containers on its network (docker_ps)
Read logs to find the real error (docker_logs)
Inspect env / mounts / health (docker_inspect)
Watch CPU/memory pressure (docker_stats)
See the network topology (docker_networks)

Then it answers from evidence, not imagination. The Fixer sees the box.

The security model

Read-only by design. The MCP tool list contains ONLY read commands — no start/stop/rm/exec. The agent can diagnose, never destroy.
Opt-in. No socket mount, no docker access. The server still runs (it reports the docker error honestly).
Scoped. Keep it on an internal network. This is virtual root: the power of sight without the power of control.

Quick start

services:
  dockroot:
    image: jpanasuk/dockroot-mcp:latest
    container_name: dockroot
    restart: unless-stopped
    networks: [your-ai-network]
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock

Connect from Hermes

# ~/.hermes/config.yaml
mcp_servers:
  dockroot:
    command: python3
    args: ["/opt/dockroot/tavern_mcp.py"]

Then ask: "what containers are running?" or "why is code-server crash-looping?" — the agent calls mcp_dockroot_docker_ps / mcp_dockroot_docker_logs and answers from the actual stack.

Or expose through MCPO (Open WebUI ecosystem)

{
  "mcpServers": {
    "dockroot": { "command": "python3", "args": ["/opt/dockroot/tavern_mcp.py"] }
  }
}

Use cases

Self-healing stacks — an agent that reads its own logs, finds the crash, applies the fix from a recipe
Onboarding — "show me everything that's running" with actual output, not a hallucinated list
Support triage — inspect logs + configs before escalating
Dev sandboxes — ephemeral stacks where the agent self-diagnoses

Devlog — 2026-08-10 · v1.0.6

Standalone MCP server: tavern connectivity toolkit + 5 read-only docker tools
Proper MCP stdio handshake (fixed the stub that sent unsolicited initialize → "Connection closed")
Static docker CLI in a slim image (Debian's docker.io package fails in slim)
Verified: hermes native client connects, MCPO connects and calls tools, docker_ps lists the live stack
Built on the basecamp adaptive discovery engine (subnet sweep, 44 fix recipes)

Links

Docker Hub · Source repo · Basecamp

Built with the Basecamp project — Hermes Agent by Nous Research under the hood.