MCP & Agent Tooling

How to Set Up Schemap MCP Server in Cursor & Claude Desktop

The Model Context Protocol (MCP) allows AI assistants like Cursor, Claude Desktop, and Windsurf to dynamically invoke tools on your local system. With Schemap MCP, your agent gains access to token-compressed schema context, automated multi-hop JOIN path solving, and pre-execution query validation.

Available Schemap MCP Tools

Tool Name Purpose
schemap_get_schema Returns token-compressed database schema with column types, foreign keys, and safety guardrails.
schemap_get_join_path Calculates the shortest verified SQL JOIN path between any two tables in the schema.
schemap_validate_query Validates SQL queries against table columns and immutability guardrails before execution.
schemap_check_blast_radius Analyzes table dependencies, cascade risks, and degree centrality.

Setup for Cursor (.cursor/mcp.json)

Create or update .cursor/mcp.json in your project repository:

{
  "mcpServers": {
    "schemap": {
      "command": "uvx",
      "args": ["schemap-tool", "mcp", "--config", "./schemap.yaml"]
    }
  }
}

Setup for Claude Desktop

Open your Claude Desktop configuration file:

{
  "mcpServers": {
    "schemap": {
      "command": "uvx",
      "args": ["schemap-tool", "mcp", "--config", "/absolute/path/to/schemap.yaml"]
    }
  }
}

Testing the Connection

Once configured, restart Cursor or Claude Desktop. You can prompt your assistant naturally:

"What is the shortest JOIN path between our customers table and the transaction line items?"

The agent will invoke schemap_get_join_path and produce a 100% verified SQL JOIN without hallucinating non-existent junction tables.

Get Started with Schemap

Install via pip or inspect the open-source repository.

Explore Schemap on GitHub