majauhar/legal-team-mcp

Multi-Agent Systems as MCP Server

This is a Model Context Protocol (MCP) server that provides a multi-agent system to help with legal document development using CAMEL-AI's workforce module. A well structured tutorial on multi-agent systems using CAMEL-AI could be followed here for additional reference. In fact, this server builds directly on top of the mentioned tutorial (also authored by me). This server used GPT-4o-mini as the LLM.

Available Tools

• process_task: Takes a task and executes it.
  • Requires task_content, task_id (optional), and additional_content (optional).
• reset: Resets the server by cleaning up the context and memory.
• get_workforce_info: Provides details about the server .
• get_children_info: Provides details of the workers (agents) of the multi-agent system.
• add_single_agent_worker: Provides the LLM (Claude) to spin more workers before starting the server.
• add_roleplaying_worker: Provides the LLM (Claude) an option to spin a role-playing worker if it considers it useful.

How to Use

Using uv

To use, we have to first install the Claude Desktop App. Once installed, open it, go to settings -> Developer -> Edit Config.

It would open a file called claude_desktop_config.json. Add the following config file there.

{
    "mcpServers": {
        "lega-team-mcp": {
            "command": "uv",
            "args": [
                "--directory",
                "/ABSOLUTE/PATH/TO/PARENT/FOLDER/legal-team-mcp",
                "run",
                "legal_team_server.py"
            ], 
            "env": {
                "OPENAI_API_KEY": "sk-proj-..."
            }
        }
    }
}

Save the file and restart the app.

Note: Sometimes, uv is not accessible globally, or it might not be in the PATH. Either install it globally or resolve the PATH issues, or put the full path to uv executable in the config key "command". This could be found using which uv on Mac/Linux, or where uv on Windows.

You could get your OPENAI API key here

The way this config would work is that uv would create a virtual environment venv and install the dependencies as specified in the repo. Upon a successful installation, the server should start successfully.

Using python

Alternatively, you could create a virtual environment yourself, install camel-ai, and then use the python interpreter to run the server, similar to what you would do on your terminal.

This would involve:

• Creating a new conda environment: conda create --name <env-name>
• Install uv by running conda install uv.
• Finally, install camel-ai by running uv pip install camel-ai

Now, find the location of python in this new environment. Following the same steps as before, add the new config (below) with suitable python location.

{
    "mcpServers": {
        "legal-team-mcp": {
            "command": "/path/to/python",
            "args": [
                "/path/to/legal_team_server.py"
            ],
            "env": {
                "OPENAI_API_KEY": "sk-proj-..."
            }
        }
    }
}