chaitanyagiri/munder-difflin
A desktop app that turns several terminal coding AIs into a coordinated office of agents
Munder Difflin is an open-source desktop app that runs multiple terminal-based coding AI tools — Claude Code, Codex, Grok, and others — at once and lets them message each other and split up work. Each AI shows up as a character on a 2D office floor, while a supervising agent called Michael assigns tasks and only asks the human for input on risky decisions. It's built with Electron and TypeScript and runs on macOS, Windows, and Linux.
What it does
- Each existing terminal AI CLI (Claude Code, Codex, Grok, Qwen, etc.) runs as its own real process in a pseudo-terminal and appears as a character on a shared office floor.
- A GOD agent (Michael) reads incoming requests, resolves routine ones on its own, and escalates only critical items like spending or destructive operations to the human.
- Agents coordinate through a local git repo acting as shared memory and mailboxes — each writes only to its own outbox/inbox, and a single router handles all git commits to avoid conflicts.
- A markdown-based memory layer with a semantic search index lets agents recall past session context in milliseconds.
- A circuit breaker automatically steers, constrains, or stops agents that loop endlessly, throw repeated errors, or exceed their budget.
Why it matters
Anyone juggling several coding-AI subscriptions manually can instead let the agents divide work among themselves, watch progress visually, and step in only when something risky comes up. Running locally for free, without needing a separate server or cloud service, also lowers the barrier to trying it.
Terms in this repo
- node-pty · a library that runs terminal programs as real processes and streams their input/output
- pseudo-terminal (PTY) · a virtual terminal that makes a program behave as if run in a real terminal
- circuit breaker · a safeguard that progressively restrains a misbehaving agent (steer, constrain, stop)
- BYOK · bring-your-own-key, using your own API credentials
- OTel (OpenTelemetry) · a standard framework for tracing and instrumenting software behavior
Repository description (English)
local multi-agent harness
Open on GitHubTrending repos
- cathrynlavery/diagram-designA skill that makes AI coding tools draw magazine-quality diagrams instead of generic rounded boxes
- public-apis/public-apisA giant crowd-curated directory of free APIs for developers
- semantica-agi/semanticaAn open-source graph infrastructure that lets AI agents show their work, not just their answers
- cactus-compute/needleA 14MB AI model small enough to run tool-calling on a phone or watch, without internet
- unslothai/unslothA desktop app that lets you run and train AI models on your own computer, no coding required
- macro-inc/macroAn all-in-one workspace where email, chat, docs, tasks, and CRM are cross-linked and share one AI memory
- harry0703/MoneyPrinterTurboAn open-source tool that turns a single topic or keyword into a finished short video, complete with script, footage, subtitles, and music
- basecamp/omarchyA ready-made, opinionated Linux setup built by DHH
Latest from METAL LAB
- Grok leaks conversations, location data via commands hidden in ciphertext
- Cursor lets agents handle long-running tasks with new "/goal" command
- NVIDIA releases Switchyard, an LLM routing proxy for coding agents
- Caveman Cuts Claude Code Token Usage by 33% Using Caveman-Speak
- Apple Keeps 2027 Target for Camera AirPods, Qwen3.8-27B Makes 80 Tool Calls Locally