Competitive Intelligence MCP Server

A Python MCP server that exposes structured competitive intelligence data as governed tools, connected to Replit Agent as the interface. A seller types a natural language prompt and gets formatted sales collateral on demand from a single source of truth.

What It Does

• Exposes two MCP tools: get_competitor_profile and list_available_profiles

• Every tool call is logged with timestamp, request ID, tool name, competitor queried, and hit/miss status

• Admin dashboard shows hits, misses, content gaps, and data freshness

• API key authentication enforced at the middleware layer before any request reaches the application

Stack

• Python 3.11

• Starlette (async web framework)

• Anthropic MCP SDK (Streamable HTTP transport)

• uvicorn (ASGI web server)

Project Structure

artifacts/
├── api-server/          # Replit routing layer (artifact.toml)
└── mcp-server/
    ├── server.py        # Entire application
    └── data/
        ├── cursor.json
        ├── lovable.json
        ├── replit.json
        └── audit.log

Running Locally

Install dependencies:

cd artifacts/mcp-server
pip install -r requirements.txt

Set environment variables:

export MCP_API_KEY=your_api_key_here
export ADMIN_KEY=your_admin_key_here

Start the server:

python server.py

Server runs on port 8000 by default.

Endpoints

MCP Tools

get_competitor_profile Fetches the full competitive intelligence profile for a named competitor. Pass the competitor name as stated by the user -- the server handles normalization.

list_available_profiles Lists all available competitor profiles with last-updated timestamps. Call this first when the competitor name is uncertain or ambiguous.

Environment Variables