LangManus is a community-driven AI automation framework designed for users who want to orchestrate complex tasks using multiple specialized AI agents. It integrates Large Language Models (LLMs) with tools for web search, crawling, and Python code execution, enabling sophisticated automation workflows.

How It Works

LangManus employs a hierarchical multi-agent system architecture. A central Coordinator handles initial task routing, delegating to a Planner for strategy creation. A Supervisor oversees specialized agents like a Researcher (for information gathering), Coder (for Python/bash execution), Browser (for web interaction), and Reporter (for summarizing results). This modular design allows for flexible task delegation and efficient execution of complex, multi-step processes.

Quick Start & Requirements

• Install: Clone the repository, create and activate a Python 3.12 virtual environment using uv, and install dependencies with uv sync.
• Prerequisites: Python 3.12, uv package manager. API keys for LLMs (e.g., Qwen, OpenAI-compatible), Tavily, and optionally Jina are required for full functionality.
• Run: uv run main.py
• Docs: LangManus English

Highlighted Details

• Supports open-source LLMs (e.g., Qwen) and OpenAI-compatible APIs.
• Integrates web search (Tavily) and neural search (Jina).
• Features a built-in Python REPL and code execution environment.
• Offers workflow graph visualization and multi-agent orchestration.

Maintenance & Community

The project emphasizes community contribution and welcomes all forms of input. Details on contributing are available in their guide.

Licensing & Compatibility

MIT License. This permissive license allows for commercial use and integration into closed-source projects.

Limitations & Caveats

Configuration requires obtaining and setting API keys for various services, which may incur costs. The project is community-driven, so the pace of development and support may vary.