Multi-robot coordination

cadenza.connect(...) wires two or more robots into a single coordination terminal so they can talk to each other and share one human goal. It reuses each robot’s user-facing narration channel for robot↔robot messages, routed through a FastMCP server under the hood.

import cadenza_lab as cadenza

go1, g1 = cadenza.go1(), cadenza.g1()

with cadenza.connect(go1, g1) as term:
    go1.comm.tell("g1", "I'll scout left, you hold position")
    g1.comm.broadcast("copy that")
    print(g1.comm.messages())

FastMCP is an optional extra. Install it with pip install "cadenza-lab[mcp]".

`connect()`

connect(*robots, narrate=True, out=None) -> CoordinationTerminal

Pass the robot controllers you want to link. The returned CoordinationTerminal is a context manager (it owns the MCP server for the session); each robot passed in gains a .comm link.

Param	Description
`*robots`	The robot controllers to connect (`go1`, `g1`, `arm`, …).
`narrate`	Echo coordination messages to the narration channel (default `True`).
`out`	Optional stream to write narration to.

term.robots         # the connected robot identifiers
term.history()      # the full message log for the session
term.close()        # tear down (handled automatically by `with`)

Robot-to-robot messaging

Each connected robot exposes a .comm link:

Method	Description
`robot.comm.tell(recipient, message)`	Send a direct message to one robot by id.
`robot.comm.broadcast(message)`	Send a message to every other connected robot.
`robot.comm.messages()`	Return the messages this robot has received.

with cadenza.connect(go1, g1) as term:
    go1.comm.tell("g1", "obstacle ahead, going around right")
    g1.comm.broadcast("holding at the doorway")
    for msg in g1.comm.messages():
        print(msg["from"], "→", msg["text"])

Splitting one goal across robots

term.coordinate(goal) takes a single human goal and splits it into per-robot subgoals, then delegates each over the MCP. Routing is by explicit target (name a robot in the goal), then capability (match the subtask to what each robot can do), then round-robin.

with cadenza.connect(go1, g1) as term:
    term.coordinate("scout the room and then both meet at the door")
    # -> go1 gets the scouting subgoal, g1 gets the hold/meet subgoal

coordinate(goal: str, *, route_by_capability: bool = True)

Set route_by_capability=False to fall back to simple round-robin assignment.

Demo: two robots, one goal

"""demo_coordination.py: connect two robots and split a goal over MCP.
Requires the mcp extra:  pip install "cadenza-lab[mcp]"
"""
import cadenza_lab as cadenza

go1, g1 = cadenza.go1(), cadenza.g1()

with cadenza.connect(go1, g1) as term:
    term.coordinate("one of you scout left while the other holds the doorway")
    go1.comm.broadcast("left flank clear")
    print("log:", term.history())

python demo_coordination.py

Drive each robot

The go1 / g1 / arm controllers each robot in the terminal drives.

Get started

Features

Tutorial

Multi-robot coordination

`connect()`

Robot-to-robot messaging

Splitting one goal across robots

Demo: two robots, one goal

Drive each robot

​connect()

​Robot-to-robot messaging

​Splitting one goal across robots

​Demo: two robots, one goal

Drive each robot

`connect()`

Robot-to-robot messaging

Splitting one goal across robots

Demo: two robots, one goal