lab-robot

Physical Execution Interface for science labs — connect AI brains to robot bodies.

Part of the LabClaw ecosystem.

Why lab-robot?

Science labs need to run experiments physically, not just plan them in software. lab-robot provides a unified driver layer that lets AI agents execute real-world lab operations — pipetting, plate handling, incubation — across different robot hardware through a single protocol. No more writing throwaway scripts for each new machine.

Features

• Unified PEI Protocol — abstract robot actions into hardware-agnostic primitives (motion, lab-ops, perception, system)
• Rich ActionResult — every action returns structured results with measurements, state changes, and error details
• Safety-First — chain-of-responsibility safety guards (force limits, workspace bounds, collision detection) at CRITICAL level by default
• Async-Native — built on Python async for concurrent robot orchestration
• Typed & Validated — full Pydantic 2.x schemas and PEP 561 type markers
• Extensible — add new robots by implementing the RobotDriver protocol

Supported Robots

Robot	Status	Mode
Opentrons OT-2	Active development	Simulate

Install

pip install lab-robot                    # core only
pip install lab-robot[opentrons]         # + Opentrons OT-2 driver

Quick Start

import asyncio
from lab_robot.types import PipetteAction
from robots.opentrons_ot2 import OT2Driver
from robots.opentrons_ot2.models import OT2DeckConfig, PipetteConfig, PipetteMount, LabwareConfig

# Configure the OT-2 deck layout
deck = OT2DeckConfig(
    slots={"1": LabwareConfig(labware_type="nest_96_wellplate_200ul_flat")},
    pipette_left=PipetteConfig(
        name="p300_single", mount=PipetteMount.LEFT, max_volume_ul=300,
        tip_rack_slots=["2"],
    ),
)

async def main():
    driver = OT2Driver(deck_config=deck, simulate=True)
    await driver.connect()
    result = await driver.execute(PipetteAction(
        volume_ul=30.0,
        source_well="A1",
        dest_well="B1",
    ))
    print(result)  # ActionResult(success=True, ...)
    await driver.disconnect()

asyncio.run(main())

Adding a New Robot

1. Copy robots/_robot_template/ to robots/<your_robot>/
2. Implement the RobotDriver protocol (connect, disconnect, execute, stop, capabilities)
3. Add your package to pyproject.toml optional dependencies

See CONTRIBUTING.md for the full guide.

Ecosystem

labclaw (orchestration)
├── lab-robot   ← you are here (physical execution)
├── device-use  (GUI & visual interaction)
└── device-skills (device drivers & skills)

lab-robot is the Physical Execution Interface (PEI) — Layer 1 of the LabClaw stack. It translates high-level lab operations into hardware-specific commands.

Architecture

See PEI Specification for the full protocol design.

Roadmap

• Phase 0 — Opentrons OT-2 simulate mode (current)
• Phase 1 — OT-2 real hardware + safety guards
• Phase 2 — Multi-robot orchestration + perception layer

Citation

If you use lab-robot in your research, please cite:

@software{labclaw_lab_robot,
  author    = {LabClaw Team},
  title     = {lab-robot: Physical Execution Interface for Science Labs},
  year      = {2026},
  url       = {https://github.com/labclaw/lab-robot},
  license   = {Apache-2.0}
}

License

Apache 2.0