Agent-to-Agent Orchestration
Agents discover peers and delegate work without being told how.
How it works
Every session gets $AMUX_SESSION (its own name) and $AMUX_URL injected at startup. A global memory file gives every agent the full REST API reference.
Key API calls
# Send a task to another session
curl -sk -X POST -H 'Content-Type: application/json' \
-d '{"text":"implement the login endpoint"}' \
$AMUX_URL/api/sessions/worker-1/send
# Claim a board task atomically
curl -sk -X POST $AMUX_URL/api/board/PROJ-5/claim
# Peek at another session's output
curl -sk "$AMUX_URL/api/sessions/worker-1/peek?lines=50"
Orchestration patterns
- Manager + workers: One agent posts tasks, workers claim and complete them
- Peer review: Agents review each other's output via peek
- Pipeline: Agent A generates code, Agent B writes tests, Agent C reviews
Get started with amux
Run dozens of Claude Code agents in parallel. Python 3 + tmux. Open source.
git clone https://github.com/mixpeek/amux && cd amux && ./install.sh
amux register myproject --dir ~/Dev/myproject --yolo
amux start myproject
amux serve # → https://localhost:8822View on GitHub