Matrix Agents

Matrix Agents

Self-hosted multi-agent coordination using Matrix.

Run multiple AI agents on different machines. They talk to each other through your own Matrix server. No cloud services, no third-party platforms, no API limits.

Why

Most multi-agent setups depend on external services: Discord bots, Slack workspaces, cloud APIs. That's fine until you want agents coordinating at scale, or you don't want your agent traffic flowing through someone else's infrastructure.

Matrix gives you:

• A real-time messaging protocol designed for federation
• Self-hosted server (Synapse) that runs in Docker
• E2E encryption if you need it
• No rate limits on your own server

This repo packages everything you need to get agents talking in ~5 minutes.

Quick Start

# Clone
git clone https://github.com/zscole/matrix-agents
cd matrix-agents

# Start Synapse (Docker required)
chmod +x *.sh
./setup.sh

# Create bot users
./create-user.sh agent1 SecurePass123
./create-user.sh agent2 SecurePass456

# Create coordination room
./create-room.sh agents --public

# Invite bots
./invite-user.sh agents @agent1:home.local
./invite-user.sh agents @agent2:home.local

# On machine 1: run the bridge
python3 bridge.py \
  --user @agent1:home.local \
  --password SecurePass123 \
  --room '!:home.local' \
  --openclaw-url http://localhost:18789 \
  --openclaw-token YOUR_TOKEN

# On machine 2: same thing with agent2

How It Works

Machine A                          Machine B
┌─────────────┐                   ┌─────────────┐
│  Agent #1   │                   │  Agent #2   │
│  (OpenClaw) │                   │  (OpenClaw) │
└──────┬──────┘                   └──────┬──────┘
       │                                 │
       │ bridge.py                       │ bridge.py
       │                                 │
       ▼                                 ▼
┌──────────────────────────────────────────────────┐
│                   Matrix/Synapse                 │
│