Repolode
Submit
Open-RUA

Open-RUA/uni-robot-agent

An open-source harness that lets LLM agents control robots through shared tools, with Codex and Claude Code runtimes and simulator integrations.

View on GitHubWebsite
Stars
258
Forks
19
Open issues
1
Last commit
yesterday

What is it?

uni-robot-agent is a Python harness for robot-use agents. It gives an agent a shared set of robot tools and feedback across several simulation environments, with native APIs handled by adapters. The package includes a CLI, a Python SDK, a local dashboard for live monitoring and replay, and integrations for RoboDojo, RoboUse and RoboLab simulators.

It supports two control modes. Direct mode has the agent call robot tools itself, while Hybrid mode adds a compatible action model such as Pi0.5. The agent runtime, model provider, tools and environment can be swapped through explicit interfaces. Depth is enabled by default in RoboDojo and can be turned off for an RGB-only observation setting.

The project targets researchers and developers who want to evaluate robot manipulation agents with open models. Real-robot integrations for ARX One and Franka are listed as coming soon.

Key features

  • 01Shared robot tool interface across simulation environments
  • 02Direct and Hybrid control modes, with Hybrid adding a Pi0.5 action model
  • 03Python SDK for running sessions and implementing custom strategies
  • 04Local web dashboard for live monitoring and replay of saved episodes
  • 05Swappable agent runtimes, model providers and environments

Use cases

  • Benchmarking robot manipulation agents on RoboDojo or RoboUse tasks
  • Testing open vision-language models for robot control with RGB-only observations
  • Building a custom control strategy on top of shared robot tools
  • Reviewing saved agent runs and tool calls through the replay dashboard

Quick start

  1. 1

    Install the package from PyPI

    pip install uni-robot-agent
  2. 2

    Prepare the RoboDojo integration

    uni-robot-agent setup --integration robodojo
  3. 3

    Set your API key and provider endpoint

    export UNI_ROBOT_AGENT_API_KEY='YOUR_API_KEY'
  4. 4

    Run a task with the local dashboard

    uni-robot-agent run --integration robodojo --deploy local --task stack_bowls --model YOUR_MODEL_ID --action-profile direct --no-sam3 --output-dir ./runs --monitor

Things to know

  • Setup is designed for a Python 3.11 environment on a Linux GPU host.
  • Running the simulator requires a model API key and a Responses-compatible provider endpoint.
  • Real-robot integrations are not yet available.

In collections

Open-source AI agent frameworks & appsRobotics & embodied AI projects

Topics

#robotics#ai-agents#llm#robot-manipulation#simulation#python-sdk#vision-language-model

FAQ

What is uni-robot-agent?
An open-source harness that lets LLM agents control robots through shared tools, with Codex and Claude Code runtimes and simulator integrations. uni-robot-agent is a Python harness for robot-use agents. It gives an agent a shared set of robot tools and feedback across several simulation environments, with native APIs handled by adapters. The package includes a CLI, a Python SDK, a local dashboard for live monitoring and replay, and integrations for RoboDojo, RoboUse and RoboLab simulators.
Is uni-robot-agent free to use? What license does it have?
uni-robot-agent is open source under the MIT license. Check the license terms before using it in commercial products.
Is uni-robot-agent still maintained?
Actively maintained. The last commit to uni-robot-agent was on Nov 6, 2026.
What can I use uni-robot-agent for?
Benchmarking robot manipulation agents on RoboDojo or RoboUse tasks Testing open vision-language models for robot control with RGB-only observations Building a custom control strategy on top of shared robot tools Reviewing saved agent runs and tool calls through the replay dashboard
How do I install uni-robot-agent?
Install the package from PyPI: pip install uni-robot-agent Prepare the RoboDojo integration: uni-robot-agent setup --integration robodojo Set your API key and provider endpoint: export UNI_ROBOT_AGENT_API_KEY='YOUR_API_KEY' Run a task with the local dashboard: uni-robot-agent run --integration robodojo --deploy local --task stack_bowls --model YOUR_MODEL_ID --action-profile direct --no-sam3 --output-dir ./runs --monitor
Where can I get uni-robot-agent?
The source code and installation instructions are on GitHub: https://github.com/Open-RUA/uni-robot-agent

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects