Postgres

Postgres MCP provides a standardized interface for AI agents to interact with PostgreSQL databases through a set of well-defined MCP commands. It supports multiple concurrent connections, basic CRUD operations, and safety and validation features to ensure secure and correct SQL execution against PostgreSQL databases.

How to use

You connect to a PostgreSQL database through an MCP client by registering a database connection, then issue high-level commands to read, insert, update, delete, or manage schema objects. The server supports both local, stdio-based usage and remote HTTP-based interaction, so you can run it as a local process or connect to a running endpoint.

How to install

cargo install postgres-mcp

Configuration and operation specifics

The MCP server can be run in two ways: as a local process using stdio, or as a remote HTTP/SSE endpoint. Use the stdio mode when running the server locally and the HTTP/SSE mode when you want to connect clients over HTTP.

Stdio (local) mode configuration example:

{
  "mcpServers": {
    "postgres": {
      "command": "postgres-mcp",
      "args": ["stdio"]
    }
  }
}

SSE/HTTP mode configuration example (URL-based):

{
  "mcpServers": {
    "postgres": {
      "url": "http://localhost:3000/sse"
    }
  }
}

Common commands and workflows

Once connected, you can perform a range of operations to manage data and schema. Typical workflows include registering a connection, running queries, and managing tables and indexes.

Register a database connection and obtain a connection ID (UUID), then use that ID for subsequent operations.

Available commands (high level)

Register a Database Connection: establish a new connection and obtain its ID for later use.

Unregister a Connection: release a previously opened connection by its ID.

Execute a SELECT Query: fetch data from a table, returning rows that match your query.

Insert Data: add new rows to a table.

Update Data: modify existing rows based on a condition.

Delete Data: remove rows from a table.

Create a Table: define a new table with columns and constraints.

Drop a Table: remove an existing table.

Create an Index: improve query performance on a column or set of columns.

Drop an Index: remove a previously created index.

Describe a Table: inspect the structure of a table, including columns and data types.

List Tables: enumerate tables within a schema.

Available tools

register

Register a new database connection and obtain a connection ID for subsequent operations.

unregister

Release an active connection using its connection ID.

query

Execute a SELECT query against a connected database and return matching rows.

insert

Insert new records into a table.

update

Update existing records based on a condition.

delete

Delete records from a table.

create_table

Create a new table with specified columns and constraints.

drop_table

Drop an existing table from the database.

create_index

Create an index to improve query performance.

drop_index

Drop an existing index.

describe_table

Describe the structure of a table, including column types and constraints.

list_tables

List all tables in a given schema.