A VS Code fork that runs Claude Code, Codex and other agent CLIs side by side, organizing work by project and git worktree feature.
madebywelch/guaca
A macOS desktop app for running groups of AI agents that chat with you, delegate work to each other, and use MCP plugins, on a local or VPS backend.
- Stars
- 54
- Forks
- 9
- Open issues
- 0
- Last commit
- 3 hours ago
What is it?
Guaca is a desktop workspace for AI agents that collaborate, write code, and use tools. The backend runs in a container, either on your own computer or on a VPS you control, and the desktop app acts as the interface to that workspace. On an always-on server, agents keep working after you close the laptop; a local workspace runs while the machine is awake.
Agents are organized into groups, each with its own agents, settings, and plugins. Plugins are MCP servers that connect agents to external services, and you decide which agents can use each plugin and its tools. Agents can talk to you and to each other, delegate tasks, and keep their own memory and working notes. Browser and computer access are available through Kernel and E2B, using the respective vendor's API keys.
The desktop app currently supports macOS and is built from source with Node.js, pnpm, Rust, and Xcode Command Line Tools. Local workspaces require Docker, while remote connections to a VPS backend do not.
Key features
- 01Groups that isolate agents, settings, and plugins from one another
- 02MCP plugin support with per-agent control over tool access
- 03Agents delegate work to each other and keep persistent memory
- 04Browser and computer access via Kernel and E2B
- 05Run the backend locally in Docker or on a remote VPS
Use cases
- Running a team of coding agents that keep working on a VPS while you are offline
- Experimenting with multi-agent workflows that share conversations and tools
- Connecting agents to your own services through MCP plugins
Quick start
- 1
Install Node.js, pnpm, Rust and Xcode Command Line Tools, then clone the repo
- 2
Build and install the desktop app from the checkout
./scripts/install.sh --no-pull --launch
Things to know
- The desktop app currently supports only macOS.
- Building from source requires Docker running for a local workspace.
- Managed compute spaces are planned for a later release and are not yet available.
In collections
Topics
FAQ
- What is guaca?
- A macOS desktop app for running groups of AI agents that chat with you, delegate work to each other, and use MCP plugins, on a local or VPS backend. Guaca is a desktop workspace for AI agents that collaborate, write code, and use tools. The backend runs in a container, either on your own computer or on a VPS you control, and the desktop app acts as the interface to that workspace. On an always-on server, agents keep working after you close the laptop; a local workspace runs while the machine is awake.
- Is guaca free to use? What license does it have?
- guaca is open source under the AGPL-3.0 license. Check the license terms before using it in commercial products.
- Is guaca still maintained?
- Actively maintained. The last commit to guaca was on Nov 7, 2026.
- What can I use guaca for?
- Running a team of coding agents that keep working on a VPS while you are offline Experimenting with multi-agent workflows that share conversations and tools Connecting agents to your own services through MCP plugins
- How do I install guaca?
- Install Node.js, pnpm, Rust and Xcode Command Line Tools, then clone the repo Build and install the desktop app from the checkout: ./scripts/install.sh --no-pull --launch
- Where can I get guaca?
- The source code and installation instructions are on GitHub: https://github.com/madebywelch/guaca
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Open-source meta-harness that orchestrates Claude Code, Codex, Cursor, and custom agents with shared policies, sandboxing, and real-time collaboration.
A desktop and web workspace where installed coding agents work as lasting, named teammates in shared conversations, with task delegation and collaborative memory.
A local-first desktop app where you build pixel-art AI agents that run real model calls and tools with your own API key.
Desktop workspace that runs many coding agents in parallel, each on its own git branch and worktree, with a live Kanban of PRs, CI and reviews.
Self-hosted TypeScript agent runtime for coordinating multiple LLM agents, with human approval gates and offline-inspectable run journals.