uvwt/agentdock
View on GitHub ↗Secure MCP runtime for AI agents to operate local machines, servers, and containers with multi-device orchestration.
296 ★53 forksGoUpdated 19d ago
What you need to know
AgentDock MCP — an independent tool runtime for AI agents that gives them secure, controlled access to every machine you operate: unified file, command, Git, Skill, MCP, browser automation and task execution across local computers, remote servers and containers, drivable from ChatGPT, Claude or Codex.
Install
Install AgentDock per Quick Start docs and connect it as an MCP server (multiple instances supported) Available via Docker Hub (agentdockio/agentdock) and GHCR (ghcr.io/uvwt/agentdock)
Usage
- •Open ChatGPT in your browser and manage multiple computers and servers from one conversation
- •Write code, run tests, change configuration and deploy in the real environment where the work belongs
- •Connect several AgentDock instances (local Mac, LAN host, cloud VPS) to coordinate multi-machine workflows
Key features
- ✓Secure, controlled file, shell, Git and MCP tool execution with explicit permission boundaries
- ✓Browser automation, Skill execution and structured, traceable, verifiable results
- ✓Multi-instance orchestration across local computers, LAN hosts and cloud servers in one conversation
- ✓No chat UI and no model inference — a focused runtime that lets agents operate real environments
Best for
Users who want their AI agents to operate real local and remote environments within permission boundaries, without consuming dedicated coding-agent quotas.
Platforms: macOS · Linux · Windows · Server · DockerClients: ChatGPT · Claude · Codex · MCP clients
Reviewed 2026-08-11
Topics
agent-skillsai-agentsautomationdevopsdockermcpmodel-context-protocolremote-controlself-hostedtool-calling
- Stars
- 296★
- Forks
- 53
- Language
- Go
- License
- Apache-2.0
- Created
- 2026-05-25
- Last push
- 2026-08-16