Agent

OpenClaw

OpenClaw, the open-source agent you run on your own machine, wired up through skills and tools to boards and bench gear.

Toolsfor OpenClaw

The most active tools first.

  1. 57 d ago

    ToolBB-0189github.com13 starsPython

    Palmimo DevKit: SDK and MCP server for a six-legged tabletop robot

    Jizai's Palmimo DevKit exposes Palmimo, a six-legged tabletop robot with 21 serial-bus servos, to Claude Code and other MCP clients. A Python SDK maps LLM tool calls to walking, waving, looking, face changes and speech. Every motion can run as a compute-only dry run first. README GIFs show the physical robot bowing and waving.

    Shown working
  2. 4 mo ago

    ToolBB-0077github.com462 starsPython

    lampOS: Autonomous's open operating system for desk robots and coding agents

    Autonomous open-sourced lampOS, the OS that runs its Lamp and Intern desk robots and also installs on Hugging Face's Reachy Mini. It connects agents such as Claude Code to the robot's cameras, microphones, speakers, motors and lights. Installable physical skills include following an object with the camera. The README's videos and photos are the vendor's own.

  3. 8 mo ago

    ToolBB-0006pypi.org

    edesto: teach coding agents to flash and test firmware

    edesto init detects the connected board, serial port and toolchain, then writes a SKILLS.md file with compile and flash commands for a coding agent. The edesto serial commands read the board's output and send it commands, and three example projects carry planted bugs for the agent to find. We have not seen it run on a board.

Builds tagged OpenClaw

Ranked by points weighed against age. New entries with early votes rise.

  1. 5 mo ago

    BuildBB-0056wired.comBuild of the week

    A Wired reporter gives his OpenClaw agent a LeRobot SO-101 arm

    Wired's Will Knight connected OpenClaw and Codex to a LeRobot SO-101 arm with a camera. Codex configured the connections, helped calibrate the joints and wrote a script that grips a red ball the camera sees. OpenClaw then guided training of a control model, checking its error rate after each run. The article's clips show the arm at work.