MCP Server for AI Agents

Integrate PostgresCompare with AI coding assistants like Claude Desktop and Claude Code using the Model Context Protocol (MCP).

PostgresCompare includes a built-in MCP (Model Context Protocol) server that allows AI coding assistants to interact with your PostgreSQL databases for schema exploration, comparison, and migration generation.

Quick Start

Start the MCP server:

pgc mcp serve

The server runs on stdio using JSON-RPC, making it compatible with any MCP client.

Claude Desktop Configuration

Add PostgresCompare to your Claude Desktop MCP configuration (claude_desktop_config.json):

{
  "mcpServers": {
    "postgres-compare": {
      "command": "pgc",
      "args": ["mcp", "serve"]
    }
  }
}

For production use with safety controls:

{
  "mcpServers": {
    "postgres-compare": {
      "command": "pgc",
      "args": ["mcp", "serve", "--read-only", "--allowed-envs", "dev,staging"]
    }
  }
}

Available Tools

The MCP server exposes 10 tools that AI agents can use:

Schema Exploration

Tool	Description
`list_environments`	List all configured database environments from `pgc.yaml`
`get_schema`	Get table, view, and function definitions for a database
`health_check`	Test database connectivity and return server version

Schema Comparison

Tool	Description
`compare_schemas`	Compare two schemas and return a diff summary
`explain_difference`	Get detailed per-property diff for a specific object
`detect_drift`	Compare current database state against a baseline snapshot

Migration & Scripting

Tool	Description
`generate_migration`	Generate a SQL migration script between two sources
`validate_sql`	Parse SQL, count statements, detect dangerous operations
`apply_migration`	Apply a SQL script to a database (with safety guards)

Snapshots

Tool	Description
`create_snapshot`	Capture a database schema to a JSON snapshot file

Safety Controls

Read-Only Mode

Prevent any write operations:

pgc mcp serve --read-only

In read-only mode, the apply_migration tool will refuse to execute any SQL.

Environment Restrictions

Limit which environments the AI can access:

pgc mcp serve --allowed-envs dev,staging

Attempts to access other environments will be rejected.

Apply Migration Guards

The apply_migration tool includes built-in safety features:

Respects environment-level readOnly: true settings in pgc.yaml
Supports maxDrops parameter to limit destructive operations
Supports maxStatements parameter to limit script size
Requires explicit confirmation for production environments

Example Interactions

Once configured, you can ask your AI assistant questions like:

“What tables are in my dev database?”
“Compare the schema between dev and production”
“Generate a migration script to sync staging with production”
“Check if there’s any schema drift from our baseline”
“Explain what changed in the users table between these two snapshots”

Data Sources

The MCP tools accept the same data source formats as the CLI:

Environment references: @dev, @production (from pgc.yaml)
Connection strings: postgres://user:pass@host:5432/db
Snapshot files: ./baseline.json
SQL files: ./schema.sql or ./schema-folder/
Git refs: git:main:schema.sql

Troubleshooting

Server won’t start

Ensure the pgc CLI is on your PATH and you have a valid license.

Environment not found

Check that your pgc.yaml configuration file exists and contains the environment definition.

Connection refused

Verify database connectivity with pgc health @env-name before using MCP.