Murage sits above your coding agents: one Chief of Staff hands work to each engine, and you can put one engine on reviewing another’s work. Open source (AGPL), on your machine, every line readable.
Free and open source. macOS, Windows and Ubuntu.
An example view. The teams are real, from the library. None of them is running yet.
Teams
One engine per window. Each coding agent is good at its own job, and none of them hands work to the others.
What did it read? When a run goes wrong, you want the exact prompt and context the agent got.
Context in one place. Memory lives inside each tool, so switching agents means explaining it all again.
Hand this off first: engine review
“Have the builder add CSV export to the reports page. When it’s done, hand it to the reviewer on Codex and tell me what it would change. Merge nothing.”
Still yours: the merge. Deleting outside a bot’s folder stops for you, even on Full access.
The jobs to hand off first. Each one ends with a draft or a question for you.
From Murage’s team library. Look inside any team before you import it, then export it once and share it with your staff.
Dev Shop
Turn requests into technical plans, review quality and operational risks, and organize implementation handoffs and release priorities.
Ship Software Safely
Get a scoped implementation plan, coordinated engineering work, and a release-readiness review backed by verification evidence.
SaaS MVP Sprint
Get an MVP specification, validation plan, landing-page copy, and visual direction.
Pre-Mortem Room
Challenge buyer appeal, channel assumptions, and potential failure points, then get a ranked list of fixes.
Why it fits this work
Murage is AGPL-3.0: read the code, fork it, contribute. In the app, see what shapes a bot shows everything a bot reads, in order, and the last turn exactly as it was sent.
Claude Code, Codex, Hermes and more, each bot on the engine that suits its job. Keep the one you use; Murage puts them on one team, you can switch any bot’s engine later, and the app is free.
Memory, files and conversations are stored on your computer, not on our servers. The AI and apps you connect receive what each job needs.
New bots start on Ask. On Full access, a bot stops before messaging anyone new, posting publicly, paying, or deleting outside its folder, unless you have allowed it for that task. Replies in an existing conversation go ahead on Auto and Full access. No limits turns the stop line off. On Hermes and other ACP engines, Ask and the stop line cover only what the engine asks Murage about: shell commands and file edits.
Claude Code, Codex, Hermes, Grok Build, Cursor, Google Antigravity, Kimi, Qwen, Droid, OpenCode, Pi, Fuigo (built in), local models (Ollama, LM Studio) and any OpenAI-compatible API (chat only). OpenClaw is coming. Codex is the official Codex CLI you install and sign in to yourself. Type / to use each engine’s own commands. To run teams and use tools, pick a tool-capable engine; a plain OpenAI-compatible endpoint works for chat and drafting only.
Yes, through MCP servers (plug-ins that give bots new tools). Custom MCP servers are saved switched off; once on, they reach every bot whose engine supports them.
Yes, each person on their own computer. Build the team once, export it, and every engineer imports the same bots, skills and routines. Logins never travel with it. Each person can message their Chief from Slack, Discord or Telegram, and bots can post drafts to your team’s Slack channels through connected apps, with your approval and up to five labeled accounts per app. Shared, always-on workspaces come with Murage Cloud, which isn’t open yet.
Download the free app, connect the AI you already pay for, and tell your Chief of Staff what needs doing. Plan on about ten minutes from install to a working team.
No account needed · Runs on the AI plan you already pay for
