MindRoom - an open-source multi-agent runtime built on Matrix

[robi]

https://github.com/mindroom-ai/mindroom

AI agents that live in your chat rooms.

MindRoom is an open-source multi-agent runtime built on Matrix that works with nearly any cloud or local AI model.

You define agents in a YAML file or in the web dashboard; MindRoom gives each one a Matrix account, and you talk to them in threads in MindRoom Chat — or any other Matrix client you already use.

Because Matrix bridges to other platforms, the same agents also work in Slack, Telegram, Discord, WhatsApp, IRC, and email — with the same persistent memory everywhere. Self-host the whole stack, or run only the MindRoom backend locally and pair it with hosted Matrix at mindroom.chat.

Features

• Multi-agent orchestration — define specialist agents and teams in config.yaml; a built-in router picks the responder when you don't @-mention one, and mentioning several agents makes them collaborate in a thread.
• Persistent memory — agents remember people, preferences, and context across conversations and platforms (Mem0 + ChromaDB, stored on your disk).
• 100+ tool integrations — Gmail, GitHub, Google Docs, Google Drive, Home Assistant, shell, Python, web search, and more, plus native Matrix tools and a per-thread todo planner, with sandboxed execution and per-tool approval rules.
• Knowledge bases (RAG) — point an agent at a folder of files; MindRoom indexes it and can watch it for changes.
• Scheduling & automation — cron or natural-language scheduled tasks (!schedule), including silent checks that post only when they find something, plus supervised background Python watchers that can call governed agent tools and wake the agent only when something changes.
• Model routing — a different model per agent, room, or thread (!model and !room_model); route sensitive rooms to local Ollama and everything else to a cloud model.
• Voice — transcription of Matrix voice messages, and text-to-speech tools via OpenAI, Groq, ElevenLabs, and Cartesia.
• Streaming responses — agents type into the room with progressive edits, visible tool traces, and cancellation.
• Plugins & hooks — drop-in plugins add custom tools, skills, and OAuth providers, and a typed event-hook system (per-hook timeouts, fault isolation) lets them observe and transform messages; reload plugins at runtime with !reload-plugins.
• Hot reload & restart-safe — config.yaml and plugin changes apply live without bringing down the stack, and conversations resume seamlessly after a restart: session history and turn state are durable on disk, so agents pick up where they left off without double-replying.
• Web dashboard — create and configure agents, teams, models, tools, credentials, and knowledge bases by clicking instead of editing YAML; chat stays in your Matrix client.
• Enterprise deployment — the same runtime scales from a laptop to multi-tenant Kubernetes with Helm charts, isolated execution workers, and egress approval for locked-down environments.

What it looks like:

You: @research @analyst @writer Create a competitive analysis report
Research: I'll gather data on our top 5 competitors...
Analyst: I'll identify strategic patterns and opportunities...
Writer: I'll compile everything into an executive summary...

Why we built this

Every AI app is a silo:

• ChatGPT knows your coding style... but can't join your team's Slack
• Claude understands your writing... but can't access your email
• GitHub Copilot helps with code... but can't see your project specs
• You teach each AI from scratch, over and over

Your human team collaborates across Slack, Discord, Telegram, and email every day — your AI should too. MindRoom agents live in one place (Matrix) and follow you everywhere via bridges, with their memory intact.

Federation even lets agents cross organization boundaries:

Your client asks in their Discord:
Client: Can our architect AI review this with your team?
You: Sure! @assistant please collaborate with them

Your Assistant: [Joins from your Matrix server]
Client's Architect AI: [Joins from their server]
Together: [They review architecture, sharing context from both organizations]

Two AI agents from different companies collaborating — impossible with app-bound assistants.