OpenManus README

The project's README from github.com/FoundationAgents/OpenManus. As one markdown file: README.md.

English | 中文 | 한국어 | 日本語

GitHub stars   License: MIT   Discord Follow Demo DOI

👋 OpenManus

Manus is incredible, but OpenManus can achieve any idea without an Invite Code 🛫!

Our team members @Xinbing Liang and @Jinyu Xiang (core authors), along with @Zhaoyang Yu, @Jiayi Zhang, and @Sirui Hong, we are from @MetaGPT. The prototype is launched within 3 hours and we are keeping building!

It's a simple implementation, so we welcome any suggestions, contributions, and feedback!

Enjoy your own agent with OpenManus!

We're also excited to introduce OpenManus-RL, an open-source project dedicated to reinforcement learning (RL)- based (such as GRPO) tuning methods for LLM agents, developed collaboratively by researchers from UIUC and OpenManus.

Project Demo

Installation

We provide two installation methods. Method 2 (using uv) is recommended for faster installation and better dependency management.

Method 1: Using conda

  1. Create a new conda environment:
conda create -n open_manus python=3.12
conda activate open_manus
  1. Clone the repository:
git clone https://github.com/FoundationAgents/OpenManus.git
cd OpenManus
  1. Install dependencies:
pip install -r requirements.txt
  1. Install uv (A fast Python package installer and resolver):
curl -LsSf https://astral.sh/uv/install.sh | sh
  1. Clone the repository:
git clone https://github.com/FoundationAgents/OpenManus.git
cd OpenManus
  1. Create a new virtual environment and activate it:
uv venv --python 3.12
source .venv/bin/activate  # On Unix/macOS
# Or on Windows:
# .venv\Scripts\activate
  1. Install dependencies:
uv pip install -r requirements.txt

Browser automation

OpenManus starts Browser Use CLI 3.0 as a default MCP server:

uvx browser-use --cli-mcp

uvx keeps Browser Use and its fast-moving dependencies isolated from the OpenManus environment. The agent receives the canonical Browser Use skill and the native browser_exec and browser_screenshot tools.

Local mode attaches to Chrome or Chromium automatically and needs no API key. For diagnostics or to install Chromium, run:

uvx browser-use --doctor
uvx browser-use install

For an isolated Browser Use Cloud browser, authenticate before starting OpenManus. The agent can then start and select a named remote browser:

export BROWSER_USE_API_KEY="bu_..."

Existing browsers can be selected with BU_CDP_URL, BU_CDP_WS, or BU_NAME. Set OPENMANUS_DISABLE_BROWSER_USE=1 to disable the default Browser Use MCP server.

BrowserGym still requires its Playwright browser:

playwright install

Configuration

OpenManus requires configuration for the LLM APIs it uses. Follow these steps to set up your configuration:

  1. Create a config.toml file in the config directory (you can copy from the example):
cp config/config.example.toml config/config.toml
  1. Edit config/config.toml to add your API keys and customize settings:
# Global LLM configuration
[llm]
model = "gpt-4o"
base_url = "https://api.openai.com/v1"
api_key = "sk-..."  # Replace with your actual API key
max_tokens = 4096
temperature = 0.0

# Optional configuration for specific LLM models
[llm.vision]
model = "gpt-4o"
base_url = "https://api.openai.com/v1"
api_key = "sk-..."  # Replace with your actual API key

Quick Start

One line for run OpenManus:

python main.py

Then input your idea via terminal!

For MCP tool version, you can run:

python run_mcp.py

For unstable multi-agent version, you also can run:

python run_flow.py

Custom Adding Multiple Agents

Currently, besides the general OpenManus Agent, we have also integrated the DataAnalysis Agent, which is suitable for data analysis and data visualization tasks. You can add this agent to run_flow in config.toml.

# Optional configuration for run-flow
[runflow]
use_data_analysis_agent = true     # Disabled by default, change to true to activate

In addition, you need to install the relevant dependencies to ensure the agent runs properly: Detailed Installation Guide

How to contribute

We welcome any friendly suggestions and helpful contributions! Just create issues or submit pull requests.

Or contact @mannaandpoem via 📧email: [email protected]

Note: Before submitting a pull request, please use the pre-commit tool to check your changes. Run pre-commit run --all-files to execute the checks.

Community Group

Join our networking group on Feishu and share your experience with other developers!

OpenManus 交流群

Star History

Star History Chart

Sponsors

Thanks to PPIO for computing source support.

PPIO: The most affordable and easily-integrated MaaS and GPU cloud solution.

Acknowledgement

Thanks to anthropic-computer-use, browser-use and crawl4ai for providing basic support for this project!

Additionally, we are grateful to AAAJ, MetaGPT, OpenHands and SWE-agent.

We also thank stepfun(阶跃星辰) for supporting our Hugging Face demo space.

OpenManus is built by contributors from MetaGPT. Huge thanks to this agent community!

Cite

@misc{openmanus2025,
  author = {Xinbing Liang and Jinyu Xiang and Zhaoyang Yu and Jiayi Zhang and Sirui Hong and Sheng Fan and Xiao Tang and Bang Liu and Yuyu Luo and Chenglin Wu},
  title = {OpenManus: An open-source framework for building general AI agents},
  year = {2025},
  publisher = {Zenodo},
  doi = {10.5281/zenodo.15186407},
  url = {https://doi.org/10.5281/zenodo.15186407},
}

OpenManus (FoundationAgents/OpenManus)

tool-use
Searching Database…

58.7K ·GitHub
OpenManus
What is this?OpenManus is an open-source general AI agent from the MetaGPT team — an open take on Manus that turns your idea into an agent working on it by itself, no invite code needed.
Do I run it myself, or is it a hosted service?You run it on your own machine: it is a Python agent application you drive from the terminal, not a hosted service.
Is there a preferred way to install it?The README documents two ways, conda and uv, and recommends uv for faster installation and better dependency management; either way you clone the repository and install the dependencies.
What does it need from me before it can actually run?An LLM API: create config/config.toml from the bundled example and put your API keys in it, since OpenManus is configured for the LLM APIs it uses.
How mature is it?The multi-agent version is explicitly labeled unstable, and the team describes the whole thing as a simple implementation — expect rough edges.
Will it run on Linux, macOS, and Windows?As a plain Python project it should run wherever Python and uv do, on any of the three, and the repository also includes a Dockerfile for container use.
How does it control a browser?It starts Browser Use CLI 3.0 as its default MCP server, so the agent gets native browser control and screenshot tools. Local mode attaches to your Chrome or Chromium automatically with no API key, and for an isolated Browser Use Cloud browser you authenticate before starting OpenManus.
What agents ship with it besides the main one?There is a DataAnalysis Agent for data analysis and data visualization tasks; you add it to the runflow in config.toml and install its relevant dependencies so it runs properly.
What is OpenManus-RL?A sibling open-source project dedicated to reinforcement-learning-based tuning methods for LLM agents (such as GRPO), developed collaboratively by researchers from UIUC and OpenManus.
How new is the project?Very new: the prototype was launched within 3 hours of the team starting, they say they are keeping building, and the GitHub repo was created in March 2025.
What are similar projects?SystemOneHarness, hive, agent-zero, turnstone, OpenDots, owl, Agentlas-OS, CUAWright
Which topics is it in?tool-use, browser-use, data-analysis, agent-framework, autonomous-agents, reinforcement-learning, browser-automation, multi-agent, framework, agent, mcp, terminal
Creator
FoundationAgents
Size
4.5 MB
Stars
58.7K
Updated
CommitAug 16Releasev0.3.0Apr 10
Age
1y
Commits · 90 Days