#77th of 9 in Agent frameworks and SDKs
openmanus
Python agent that carries out terminal-entered tasks with LLM tools and a browser
- Stars
- 58.6k
- License
- MIT
- Last commit
- Aug 2026
- Last release
- Apr 2025
- Language
- Python
54,490 on 24 Mar 202658,638 on 13 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
OpenManus is a Python agent framework, an open reimplementation of Manus, that takes a task typed in the terminal and carries it out with tools. It drives a browser through the Browser Use CLI MCP server and calls an LLM endpoint configured in config.toml. Extra entry points cover an MCP tool version, an experimental multi-agent flow, and an optional data analysis agent.
Who it is for: Developers wanting a hackable open Manus-style agent run from the terminal
Strengths
- Browser automation via Browser Use MCP server; local Chrome mode needs no API key
- Separate entry points: main.py, run_mcp.py and run_flow.py for multi-agent
- Simple config.toml with separate LLM and vision model settings
- MIT license, active commits, and a Hugging Face demo space
Weaknesses
- Multi-agent flow is described by the README as unstable
- Terminal-only interface, no web UI or user accounts
- No tagged release since 2025-04-10
- README shows only OpenAI-style API setup, no local model guidance
What it needs
- no GPU
- Docker
- Needs Python 3.12, uv or conda, LLM API key, Browser Use CLI (uvx), Chrome or Chromium, Playwright (BrowserGym)
- Models: OpenAI-compatible API (gpt-4o in example config)
Also in Agent frameworks and SDKs
See all 9| Rank | Project | Score |
|---|---|---|
| 1 | langchainPython framework for building LLM agents and apps from interoperable components | 88 out of 100 |
| 2 | browser-usePython library that lets an LLM agent control a web browser | 87 out of 100 |
| 3 | crewaiPython framework for multi-agent workflows with role-based Crews and event-driven Flows | 77 out of 100 |
| #4 | agnoPython SDK and runtime for building and serving agents behind a REST API | 69 out of 100 |
| #5 | llama_indexPython framework for RAG and agent apps over your own data | 68 out of 100 |