OpenManus

An open-source, flexible, and accessible alternative to Manus AI.

What is OpenManus?

OpenManus is an incredible, open-source AI agent that aims to achieve any idea without the need for an invite code. Built collaboratively by researchers from MetaGPT and UIUC, the initial prototype was launched in just 3 hours and is continuously being improved. It provides a simple yet powerful implementation of an AI agent, welcoming suggestions, contributions, and feedback from the community.

Key Innovation: Open-source accessibility with no invite codes required, coupled with an active community and ongoing development.

No Invite Code

Accessible to everyone immediately.

Fast Prototyping

Initial version built in just 3 hours.

Open Source

Community-driven development.

MetaGPT Origins

Built by core authors from MetaGPT.

Methods & RL Tuning

Alongside the main agent, the team introduced OpenManus-RL. This is an open-source project dedicated to implementing reinforcement learning (RL)-based tuning methods for LLM agents.

GRPO Tuning

OpenManus-RL actively utilizes advanced reinforcement learning methods such as Group Relative Policy Optimization (GRPO) to enhance the reasoning and execution capabilities of the agent.

Multi-Agent Support

Beyond the general OpenManus Agent, it integrates specialized agents like the DataAnalysis Agent, specifically tailored for data manipulation and visualization tasks using `run_flow`.

Installation Guide

OpenManus provides two primary methods for installation. Using `uv` is highly recommended for faster installation and superior dependency management.

Method 1: Using uv (Recommended)

Install uv (fast Python package installer)
Clone the repository
Create a virtual environment: uv venv --python 3.12
Activate it and install: uv pip install -r requirements.txt

curl -LsSf https://astral.sh/uv/install.sh | sh
git clone https://github.com/FoundationAgents/OpenManus.git
cd OpenManus
uv venv --python 3.12
source .venv/bin/activate
uv pip install -r requirements.txt

Method 2: Using Conda

Create a conda environment: python=3.12
Activate the environment
Clone the repository
Install dependencies using pip

conda create -n open_manus python=3.12
conda activate open_manus
git clone https://github.com/FoundationAgents/OpenManus.git
cd OpenManus
pip install -r requirements.txt

Browser Automation (Optional): If your use case requires browser automation, run playwright install after installing dependencies.

Configuration & Quick Start

1. Configuration

OpenManus requires configuration for the LLM APIs it uses. Create a config.toml file in the config/ directory.

cp config/config.example.toml config/config.toml

# Example config.toml modifications
[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

2. Running the Agent

You can run OpenManus in several modes depending on your needs:

python main.py Standard mode. After running, input your idea via the terminal.
python run_mcp.py MCP tool version.
python run_flow.py Multi-agent version. Requires adding use_data_analysis_agent = true under [runflow] in your config.