Munder Difflin is a multi-agent harness that coordinates multiple AI coding agents on your local machine.
The tool solves the problem of running multiple AI agents efficiently within the constraints of existing API subscriptions and hourly rate limits. It wraps several coding-focused language models—Claude Code, Gemini, OpenAI Codex, xAI Grok, Kimi Code, Qwen, OpenCode, Crush, pi.dev, and GitHub Copilot CLI—allowing you to bring your own API keys or use local models. The approach works by converting terminal-based coding CLIs into coordinated agents that persist while you're away. Each agent receives long-term memory and a mailbox for communication. A central coordinator agent called Michael routes work between them, visualizing the office as a 2D floor where agents appear as avatars working on shared tasks.
Developers should choose this tool if they want to automate coding work across multiple AI models without paying for dedicated agent infrastructure, and if they already subscribe to the underlying services. It suits projects where you need parallel work streams coordinated intelligently—the tool handles routing and memory management so agents can build context over time. The project positions itself as free and performant, working within the hourly limits of subscriptions you already pay for, distinguishing it from solutions that require separate agent platform costs.
The project is in working prototype status and actively welcomes pull requests. Development spans a full stack built with Electron, React, TypeScript, Pixi.js for visualization, xterm.js for terminal rendering, and node-pty for process management, indicating sustained effort across frontend, backend, and system integration layers.
