Smith_Agentic_MCP

MCP server for Smith_Agentic. Exposes Smith_Agentic crew runs as tools for Claude Code and Claude Desktop via the Model Context Protocol.

Communicates with Smith_Agentic's FastAPI server over HTTP. No dependency on crewai, Ollama, or any Smith_Agentic internals - just two packages: mcp and httpx.

Prerequisites

Smith_Agentic must be running before this MCP server can do anything:

cd Smith_Agentic
python ui/server.py
# Smith_Agentic API now available at http://localhost:8765

Installation

git clone https://github.com/AddisonTech/Smith_Agentic_MCP.git
cd Smith_Agentic_MCP
pip install -r requirements.txt

Setup

Claude Code

claude mcp add smith_agentic python C:/path/to/Smith_Agentic_MCP/server.py

Verify it loaded:

/mcp

Claude Desktop

Edit %APPDATA%\Claude\claude_desktop_config.json (Windows) or ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{
  "mcpServers": {
    "smith_agentic": {
      "command": "python",
      "args": ["C:/path/to/Smith_Agentic_MCP/server.py"]
    }
  }
}

Restart Claude Desktop after saving.

Virtual environment: If your dependencies are in a .venv, point at that Python directly:

"command": "C:/path/to/Smith_Agentic_MCP/.venv/Scripts/python.exe"

Configuration

| Variable | Default | Description | |---|---|---| | SMITH_AGENTIC_URL | http://localhost:8765 | Base URL for the Smith_Agentic API server |

Set it if Smith_Agentic is running on a non-default port or a remote host:

# Claude Code - set in your shell profile or pass inline
SMITH_AGENTIC_URL=http://192.168.1.10:8765 python server.py

Or in Claude Desktop config:

{
  "mcpServers": {
    "smith_agentic": {
      "command": "python",
      "args": ["C:/path/to/Smith_Agentic_MCP/server.py"],
      "env": {
        "SMITH_AGENTIC_URL": "http://192.168.1.10:8765"
      }
    }
  }
}

Available Tools

| Tool | What it does | |---|---| | check_smith_agentic | Verify Smith_Agentic is running and Ollama is reachable | | list_crews | Show available crews and their default models | | list_models | Show installed Ollama models | | run_crew(goal, crew, model, chain) | Start a crew run, returns run_id immediately | | get_run_status(run_id) | Poll a run for status and recent output | | list_output_files | List all files in Smith_Agentic's outputs/ | | read_output_file(path) | Read a file from outputs/ (supports subdirectory paths) |

Usage

Once connected, you can ask Claude naturally:

"Check if Smith_Agentic is running, then start a default crew run: write a Python class for managing a connection pool with retry logic."

Claude will:

1. Call check_smith_agentic to confirm the system is ready
2. Call run_crew to start the run and get a run_id
3. Call get_run_status every 30-60 seconds until status is completed
4. Call list_output_files to see what was produced
5. Call read_output_file to show you the result

Crew options

| Crew | Use for | |---|---| | default | General software tasks - specs, code, analysis | | plc | Rockwell/Allen-Bradley Logix PLC programs | | react | Industrial React/MUI HMI components | | vision | Vision_Inspect defect analysis (requires Vision_Inspect on port 8000) | | safety | Validate an existing deliverable - QA, security, deploy checks | | ops | Generate docs and telemetry for an existing deliverable |

Chain flag

Pass chain=True to automatically run the safety and ops crews after the primary crew finishes:

"Run a default crew on this goal with chain=True: build a Python rate limiter."

The run will execute: default → safety → ops, producing the full set of outputs including QA report, security report, deploy verdict, docs, and telemetry.

Project structure

Smith_Agentic_MCP/
├── server.py          # MCP server - all tools defined here
├── requirements.txt   # mcp + httpx only
└── README.md