Getting Started

Prerequisites

• .NET 10 SDK
• PostgreSQL with Apache AGE extension
• Optional: pgvector extension (for semantic search)

1. Start the Database

The included Docker setup runs PostgreSQL 17 + AGE 1.6.0 + pgvector:

git clone https://github.com/Neftedollar/age-mcp.git
cd age-mcp
docker compose up -d

This creates a database on port 5435 with user agemcp / password agemcp.

To customize versions or port:

PG_MAJOR=16 AGE_VERSION=1.5.0 DB_PORT=5440 docker compose up -d

See Configuration for all options.

2. Install age-mcp

As a global dotnet tool (recommended)

dotnet tool install --global AgeMcp

Then run:

export AGE_CONNECTION_STRING="Host=localhost;Port=5435;Database=agemcp;Username=agemcp;Password=agemcp"
age-mcp

From source

cd age-mcp
dotnet run

As a local tool

dotnet new tool-manifest
dotnet tool install --local AgeMcp
dotnet age-mcp

3. Connect to Claude

Claude Desktop

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

{
  "mcpServers": {
    "age-mcp": {
      "type": "stdio",
      "command": "age-mcp",
      "env": {
        "AGE_CONNECTION_STRING": "Host=localhost;Port=5435;Database=agemcp;Username=agemcp;Password=agemcp",
        "TENANT_ID": "default"
      }
    }
  }
}

Claude Code

Add to .claude/settings.json or configure via /mcp:

{
  "mcpServers": {
    "age-mcp": {
      "type": "stdio",
      "command": "age-mcp",
      "env": {
        "AGE_CONNECTION_STRING": "Host=localhost;Port=5435;Database=agemcp;Username=agemcp;Password=agemcp",
        "TENANT_ID": "default"
      }
    }
  }
}

4. Try It

Once connected, ask your AI assistant:

“Create a graph called ‘team’ and add three people: Alice (engineer), Bob (designer), Carol (PM). Connect Alice and Bob with a WORKS_WITH edge.”

The assistant will call get_or_create_graph, upsert_vertex (3x), and upsert_edge automatically.

“Show me the schema of the team graph”

Calls get_schema – returns label counts.

“Export the team graph as JSON”

Calls export_graph – returns all vertices and edges.

“Generate a visualization of the team graph”

Calls generate_visualization – returns standalone HTML with an interactive vis.js network diagram.