Cross-tool

Give Your AI Agent a Postgres Database Over MCP (Free)

Every Hatchable project comes with its own private Postgres. Connect your agent once and it can create tables, run SQL and write migrations over MCP, then ship the app that uses them. Nothing to provision.

What happens next

Step 1

Connect your agent

Add https://hatchable.com/mcp to Claude Code, Cursor, Codex, ChatGPT or Gemini. Approve the sign-in once; that creates your free account.

Step 2

Ask for the data model

Say what you need: "Create a projects table with a name, status and due date, and a tasks table that belongs to it." The agent writes the migration and runs it. You can open the project's database view in the console and see the rows.

Step 3

Ship the app on top

The same agent writes the pages and API routes that read and write those tables, adds sign-in if the data is per-user, and deploys. Live at your-app.hatchable.site.

The MCP server URL. Paste it into Claude Code, Cursor, Codex, ChatGPT or Gemini; OAuth does the rest
https://hatchable.com/mcp

What the agent can do with the database over MCP

  • Create tables, run queries and write migrations against a standard PostgreSQL database that belongs to your project alone.
  • Inspect the schema before it changes anything, and explain what it plans to do.
  • Scope rows to signed-in users with one line of config when the app is multi-user.
  • Keep going after launch: every later change is a sentence to the same agent.

How this compares with a database MCP server

A database MCP server (Supabase's, a community Postgres server, a SQL server) typically gives an agent access to a database you provisioned somewhere else: you create the project, copy keys or a connection string, and wire the server into each client. That is the right tool when the database already exists. Here the database, the hosting, sign-in, storage and email are one thing and exist the moment the project does, so there is nothing to provision or paste. General characterisation as of August 2026; check each vendor for current details.

Limits, stated plainly

The free plan includes unlimited private projects and one published app, with a monthly database allotment sized for personal projects and prototypes; Builder ($12 a month) is roughly 5x and each extra $12 block multiplies it again. Every project on every plan gets its own private database; there is no shared data access between projects or accounts. The full table is on /pricing.

Frequently asked

Is it real PostgreSQL?
Yes. Standard PostgreSQL with real SQL and real migrations, one private database per project.
How do I see the data?
From your agent over MCP, from the console's database view (rows, tables, CRUD), and from your app's own code. You can also download the data as SQL.
Is my data isolated from other projects?
Yes. Every project gets its own private database and there is no shared data access between projects or accounts, on every plan.
Does this replace a Supabase MCP server?
If your goal is a database the agent can use for the app it is building, yes: it is already here. If you have an existing database elsewhere, keep using a database MCP server for that.
Which agents work?
Anything that speaks MCP: Claude Code, Claude, Cursor, Codex, ChatGPT, Gemini, Antigravity, Cowork. The URL is the same for all of them.

Ready to ship?

Every Hatchable project comes with its own private Postgres. Connect your agent once and it can create tables, run SQL and write migrations over MCP, then ship the app that uses them. Nothing to provision.

Connect your agent free