PostgreSQL MCP

What it does

PostgreSQL MCP connects Claude to any Postgres database using a standard connection string. Claude becomes a capable SQL collaborator: it can explore an unfamiliar schema, write complex analytical queries, identify performance bottlenecks using EXPLAIN, and manage data with INSERT and UPDATE statements.

It works with local Postgres instances, cloud-hosted databases (AWS RDS, GCP Cloud SQL, Azure Database, Supabase, Neon), and any Postgres-compatible system like CockroachDB or Aurora.

• Run any SQL query with results formatted as a table
• List all tables across all schemas
• Describe column definitions, types, nullability, and defaults
• List available schemas within the database
• Run EXPLAIN ANALYZE to profile query performance
• Execute multi-statement transactions

How to connect

Replace the connection string with your actual database credentials:

claude mcp add postgres -- npx -y @modelcontextprotocol/server-postgres \
  "postgresql://myuser:mypassword@localhost:5432/mydb"

For cloud databases, use the full connection URL from your provider's dashboard.

Available tools

• query — run a SQL query and return results
• list_tables — list tables in the current schema
• describe_table — return column definitions for a table
• list_schemas — list all schemas in the database
• explain_query — return the query execution plan
• execute_transaction — run multiple statements atomically

Example usage

Ask Claude to investigate slow queries or build a monthly cohort analysis:

# User prompt
Find the 5 slowest queries in my orders table,
show me their execution plans, and suggest
which columns to index.

list_tables()
describe_table("orders")
explain_query("SELECT * FROM orders
  WHERE customer_id = $1 AND status = 'pending'
  ORDER BY created_at DESC LIMIT 20")
→ Seq Scan cost=0.00..8243.50
→ Claude recommends: CREATE INDEX ON orders(customer_id, status)