YellowbrickData/YellowbrickMCP

Yellowbrick MCP Server

Yellowbrick MCP Server provides a seamless interface to query your Yellowbrick database using natural language, with optional document search powered by OpenAI embeddings.

---

Table of Contents

• Getting Started
• Running with Claude
• Running with VS Code
• Configuration
• Example Questions

---

Getting Started

1. Set your environment variables:
   • YELLOWBRICK_CONNECTION (your Yellowbrick DB connection string)

• OPENAI_API_KEY (required for document search)
• EMBEDDING_TABLE (optional, for vector document search)

2. Build the Docker image:

   source build.src
   

---

Running with Claude

1. Build the Docker image as above.

2. Add the following segment to your Claude config at: /Users/[Your Username]/Library/Application Support/Claude/claude_desktop_config.json

   {
     "globalShortcut": "",
     "mcpServers": {
       "Yellowbrick": {
         "command": "docker",
         "args": [
           "run", "-i", "--rm",
           "-e", "YELLOWBRICK_CONNECTION",
           "-e", "OPENAI_API_KEY",
           "-e", "EMBEDDING_TABLE",
           "ybmcp"
         ],
         "env": {
           "YELLOWBRICK_CONNECTION": "postgresql://[user]:[pwd]@[host]:[port]/[dbname]",
           "OPENAI_API_KEY": "sk-XXXXXXXXX",
           "EMBEDDING_TABLE": "my_embeddings"
         }
       }
     }
   }
   

3. Restart Claude.

4. If you have the PremDB schema loaded (from samples), try these queries:

   • "How many matches ended in a tie?"
   • "What is the win/loss record for each team?"
   • "describe the match table?"
   • "How many rows in the teams table?"

---

Running with VS Code

1. Press CMD+SHIFT+P, type MCP: Add, and choose stdio.

2. Enter the following command (replace placeholders with your credentials):

   docker run -i --rm \
     -e YELLOWBRICK_CONNECTION=postgresql://[user]:[pwd]@[host]:[port]/[dbname] \
     -e OPENAI_API_KEY=sk-XXXXXXXX \
     -e EMBEDDING_TABLE=mcpvdb2 \
     -e READ_ONLY=true \
     ybmcp
   

---

Configuration

• YELLOWBRICK_CONNECTION: Your Yellowbrick PostgreSQL connection string.
• OPENAI_API_KEY: Your OpenAI API key (required for document search).
• EMBEDDING_TABLE: Name of the table for vector embeddings (optional).
• READ_ONLY: Set to true (default) to disable write operations (INSERT, UPDATE, DELETE, etc.) for safety. Set to false to allow write queries.

---

Example Questions

If you have the PremDB schema loaded (from samples), try these queries:

• "How many matches ended in a tie?"
• "What is the win/loss record for each team?"
• "describe the match table?"
• "How many rows in the teams table?"

---

For more details, see the comments in mcp.json and the Dockerfile.