human-mcp
MCP server that provides humans as MCP tools
overview
human-mcp is an MCP server that allows AI assistants to leverage human capabilities: it receives requests from AI assistants, displays instructions to humans, and returns responses from humans to the AI assistant.
Key features:
- Accepts tool execution requests (via STDIN) from MCP clients
- Write the instructions required for execution to a SQLite database
- The Streamlit application monitors SQLite, displays instructions to the human, and prompts for responses.
- Write the results of human input via Streamlit to SQLite
- The MCP server reads the results from SQLite and returns them to the client (via STDOUT) as an MCP response.
- human_eye_tool : A human eye is used to describe a situation or locate something specific.
- human_hand_tool : A human using his or her hand to perform a simple physical manipulation.
- human_mouth_tool : A human uses his mouth to say the specified words.
- human_weather_tool : A human checks and reports the weather in your location.
- human_ear_tool : A human uses his ears to hear sounds and describe the situation.
- human_nose_tool : A human uses their nose to identify smells.
- human_taste_tool : A human uses his mouth to taste food and describe its taste.
set up
Prerequisites
- Python 3.12 or higher
- uv
- SQLite3
Installation Instructions
- Clone the repository
git clone https://github.com/yourusername/human-mcp.git
cd human-mcp
- Create and activate the virtual environment
uv venv
source .venv/bin/activate
- Install dependencies
How to use
- Install MCP server
- Connect to MCP server from Claude
"human-mcp": {
"command": "uv",
"args": [
"run",
"--with",
"mcp[cli]",
"mcp",
"run",
"$PATH_TO_REPOSITORY/human_mcp/mcp_server.py"
]
}
- Launch Streamlit UI in a second terminal
- Access the Streamlit UI in your browser (usually http://localhost:8501 )
- Once you submit your request through your MCP client (e.g. Claude Desktop), the task will appear in the Streamlit UI.
- Once you enter your response in the Streamlit UI and click the "Send Response" button, the response will be sent back to the MCP client.
Project Structure
human-mcp/
├── human_mcp/ # メインのPythonパッケージ
│ ├── __init__.py # パッケージマーカー
│ ├── db_utils.py # SQLite関連ユーティリティ
│ ├── tools.py # ツール定義
│ ├── mcp_server.py # MCPサーバー本体
│ └── streamlit_app.py # Streamlit UI アプリ
├── human_tasks.db # SQLite データベースファイル (実行時に生成)
├── pyproject.toml # プロジェクト設定、依存関係
└── README.md # このファイル
license
MIT
Notes
This project is intended for use as a joke. In actual operation, it is necessary to take into account the burden on human operators and response delays.