Lazy Manus

An opensource Manus-like agent using lazyllm(https://github.com/LazyAGI/LazyLLM)

Current Func

1. Plan and give Todo list for input task.
2. So many tools supported(Bocha AI search, browser use, coding, file system management, Amap func...)
3. MCP server supported.

Plan-and-Solve Structure

Quick Start

1. set deepseek api key "export LAZYLLM_DEEPSEEK_API_KEY=xxx"
2. set Bocha AI search api key in .env
3. Depends on your system(Linux/Windows), set specific command in ./mcp_config_list.json for every mcp server.
4. set api keys for github and amap mcp servers in mcp_config_list.json
5. start playwright mcp server in terminal and set url for remote_playwright mcp server in mcp_config_list.json.(e.x. For Windows: lazyllm deploy mcp_server --sse-port 11287 cmd -- /c npx @playwright/mcp@latest)
6. run python -m run_manus.py

MCP Server List

1. filesystem: https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem
2. github: https://github.com/modelcontextprotocol/servers/tree/main/src/github
3. playwright: https://github.com/microsoft/playwright-mcp
4. amap https://mcp.so/zh/server/amap-maps/amap

Attention!!!

Friendly reminder: Running the application may consumes a significant amount of token quota, take care :)

TODO List

1. Replay mode