other / agent
OpenManus
A Python agent you run from the terminal after adding LLM API settings, with Browser Use tools, an MCP tool version and a multi-agent version.
At a glance
- Purpose
- A Python agent you run from the terminal after adding LLM API settings, with Browser Use tools, an MCP tool version and a multi-agent version.
- Runs in
- Local agent with shell, MCP server
- Operating systems
- macOS, Windows
- Needs
- Runtime: Python 3.12; Needs internet: Not stated; Needs accounts or API keys beyond your agent: Not stated
- Files and data
- Reads files: Not stated; Writes files: Not stated; Sends data to other services: Not stated
- Tool type
- Agent
- Delivery
- From GitHub
- Version
- Latest on the main branch
- Access
- Code on GitHub
- Compatibility
- According to the README, OpenManus installs with conda or uv into a Python 3.12 environment. It shows environment activation for Unix and macOS and for Windows. The example configuration sets a model name, a base URL and an API key for an LLM API. The README says local browser mode attaches to Chrome or Chromium and describes authenticating a Browser Use Cloud browser with BROWSER_USE_API_KEY. The README also names MCP, BrowserGym and Playwright.
- External costs
- According to the README, OpenManus requires configuration for the LLM APIs it uses. Its example configuration sets a model (gpt-4o), a base URL and an api_key. The README says local browser mode needs no API key and that an isolated Browser Use Cloud browser needs BROWSER_USE_API_KEY. The README states no prices. The providers you connect may charge for usage.
- Permissions
- The README shows installing dependencies with pip or uv and installing uv through a downloaded script piped into sh. The README shows creating config/config.toml from an example and editing it to add LLM API keys and settings. The README says OpenManus starts Browser Use CLI 3.0 through uvx as a default MCP server and gives the agent browser_exec and browser_screenshot tools. The README says local mode attaches to Chrome or Chromium automatically, and an existing browser can be selected with BU_CDP_URL, BU_CDP_WS or BU_NAME. The README says OPENMANUS_DISABLE_BROWSER_USE=1 disables that default server. The README describes Chromium and Playwright browser installation commands and remote Browser Use Cloud access authenticated through BROWSER_USE_API_KEY. The README does not describe approval or sandbox modes.
- Usage terms
- MIT License, as stated in the repository. See its LICENSE file for the exact terms.
Code on GitHub
FoundationAgents/OpenManus
MITNo fortress, purely open ground. OpenManus is Coming.
Public GitHub data fetched · 0 days old
View on GitHub ↗What it does
According to the README, OpenManus is an agent project run from the command line. The README describes configuring LLM API settings, starting a Python script and entering an idea at the terminal. The README calls the project a simple implementation.The README says OpenManus starts Browser Use CLI 3.0 as a default MCP server and supplies the agent with the Browser Use skill, browser_exec and browser_screenshot tools. The README describes an MCP tool version and an unstable multi-agent version. The README describes a DataAnalysis agent for data analysis and visualization, disabled by default and requiring additional dependencies.
The README links Chinese, Korean and Japanese versions and describes the separate OpenManus-RL project as focused on reinforcement-learning-based tuning methods for LLM agents.
Who it is for
People who want to run or build on an open-source general AI agent from the terminal with their own LLM API settings.Example tasks
Run python main.py and type an idea at the terminal.Run the MCP tool version with python run_mcp.py.
Run the multi-agent version with python run_flow.py.
Turn on the DataAnalysis agent for data analysis and data visualization by setting use_data_analysis_agent to true in config.toml.
Let the agent use a local Chrome or Chromium browser through the default Browser Use MCP server.
Getting started
According to the README, there are two install methods. Method 1 creates a conda environment with Python 3.12, clones the repository and runs pip install -r requirements.txt. Method 2, which the README recommends, installs uv, clones the repository, creates a virtual environment with uv venv --python 3.12, activates it and runs uv pip install -r requirements.txt. Then run cp config/config.example.toml config/config.toml, add your LLM API settings in that file, and start OpenManus with python main.py. The README has the exact commands, including the optional Browser Use setup.Limitations
The README requires configuration of LLM APIs in config/config.toml. The README's installation methods create Python 3.12 environments and show activation on Unix/macOS and Windows. The README says Browser Use runs as a default MCP server, local mode attaches to Chrome or Chromium, and Chromium can be installed through Browser Use. The README says BrowserGym requires its Playwright browser. The README calls the multi-agent version unstable and says the disabled-by-default DataAnalysis agent needs additional dependencies. The README calls OpenManus a simple implementation and does not describe approval or sandbox modes.The repository can change at any time; check its README before you install.