Skip to content

Releases

Embodify 0.1.0a2

Embodify 0.1.0a2 Pre-release
Pre-release

Choose a tag to compare

@YidaYang YidaYang released this 02 Oct 09:28

One plugin: installing Embodify now gives your agent both the MCP server and the skills.

Get started

Paste this into your agent:

Install Embodify (https://github.com/YidaYang/embodify) for yourself, then set up a LIBERO environment or connect to my existing simulator.

Or install the plugin yourself (needs uv):

# Claude Code
claude plugin marketplace add YidaYang/embodify
claude plugin install embodify@embodify

# Codex
codex plugin marketplace add YidaYang/embodify
codex plugin add embodify@embodify

What's new

  • One plugin. The plugin now brings the MCP server as well as the skills, and is renamed from embodify-skills to embodify. The server starts with uvx straight from PyPI, so there is nothing else to install.
  • Settings file. The server reads its options from ~/.embodify/config.json (--config), and runs on the Fake backend until the file exists. Connecting a simulator now means writing that file instead of registering the server again.
  • Run logs in your home folder. Logs go to ~/.embodify/runs by default, and the monitor reads them from there.
  • On PyPI and the MCP Registry. pip install embodify-mcp (or the alias embodify), and io.github.YidaYang/embodify in the MCP Registry.

Upgrading from 0.1.0a1: uninstall the embodify-skills plugin and remove the embodify server you registered by hand, then install the embodify plugin. Move your server options into ~/.embodify/config.json.

Full changelog: CHANGELOG.md

Embodify 0.1.0a1

Embodify 0.1.0a1 Pre-release
Pre-release

Choose a tag to compare

@YidaYang YidaYang released this 02 Oct 08:37

The first release of Embodify: give your agent a body.

Embodify lets the agent you use every day, such as Claude Code, Codex or any other MCP-capable agent, see and control robots directly, while keeping its memory, its tools and everything it knows about you.

Get started

Paste this into your agent:

Install Embodify (https://github.com/YidaYang/embodify) for yourself, then set up a LIBERO environment or connect to my existing simulator.

Or install by hand:

python -m pip install embodify-mcp
embodify-mcp-smoke

What's in 0.1.0a1

MCP server (embodify-mcp)

  • Tools: get_session_info, list_tasks, reset_task, observe, move_relative, set_gripper, stop_episode, and control_arms for coordinated two-arm motion.
  • Backends: one- and two-arm Fake, LIBERO with all 130 tasks, and RoboDojo with dual ARX X5 arms on all 54 simulation tasks.
  • Remote backends over SSH or TCP, with heartbeats for slow links and reconnection through reset_task.
  • Live monitor and replay in the browser, in English or Chinese.
  • MCP over stdio, protocol versions 2024-11-05 to 2025-11-25.

Skills pack (embodify-skills)

  • embodied-control, robot-profile and task-experience.
  • Coming soon: object-segmentation and depth-ranging.

Installation

  • Claude Code: claude plugin marketplace add YidaYang/embodify, then claude plugin install embodify-skills@embodify.
  • Codex: codex plugin marketplace add YidaYang/embodify, then codex plugin add embodify-skills@embodify.
  • Register the MCP server as embodify; see the README.

Coming next: RoboTwin and the LeRobot SO-101 arm.

Full changelog: CHANGELOG.md