PostgREST is a Haskell-written server that serves a fully RESTful API from any existing PostgreSQL database, and it is aimed at developers, database administrators, and operators who want a standards-compliant HTTP API without writing and maintaining one from scratch.

What it is

PostgREST is a standalone HTTP server, written in Haskell with the Warp HTTP server, that exposes an existing PostgreSQL database as a RESTful API. It uses the Hasql library to keep a pool of database connections, speak the PostgreSQL binary protocol, and remain stateless so the server can scale horizontally. Authentication is handled via JSON Web Tokens, and authorization is delegated to role information defined inside the database. The server assumes the identity of the authenticated user for the duration of the connection, so it cannot do anything that user could not do. Versioning is performed through database schemas, and OpenAPI output provides self-documentation that tools such as Swagger-UI can render.

The concrete problem it solves is the custom API layer that teams otherwise write, maintain, and secure by hand. Rather than relying on an Object Relational Mapper and custom imperative coding, PostgREST requires declarative constraints directly in the database and delegates as much calculation as possible to PostgreSQL: serializing JSON responses in SQL, validating data, enforcing authorization, combining row counting and retrieval, and posting data in a single command with returning *. The README describes the result as a cleaner, more standards-compliant, faster API than one written from scratch. It lives in the PostgreSQL ecosystem and replaces the hand-written REST API and ORM-centric data access layer that would otherwise sit between a database and its clients.

Key capabilities

• Serves a fully RESTful API from any existing PostgreSQL database without application server code.
• Authenticates with JSON Web Tokens (JWT) and delegates authorization to role information defined in the database.
• Generates OpenAPI documentation for the live API, which can be rendered with Swagger-UI.
• Reports and limits the number of rows returned by an endpoint through range headers.
• Versions APIs through database schemas, allowing tables and views to be exposed while underlying tables are hidden behind public-facing views.
• Delegates JSON serialization, data validation, authorization, combined row counting and retrieval, and data post in a single command (returning *) to SQL.
• Uses Hasql with a database connection pool, the PostgreSQL binary protocol, and a stateless design for horizontal scaling.

Who uses it and how

• Teams that already run PostgreSQL and need an HTTP API over existing tables and views, without writing and maintaining separate API code.
• Operators deploying the server on a supported platform or through the Docker image postgrest/postgrest.
• Applications that need JWT-based authentication while keeping authorization rules in PostgreSQL roles.
• Developers who want interactive, always-current API documentation through OpenAPI and Swagger-UI.
• Self-hosters and horizontally scaled deployments; the README reports subsecond response times for up to 2000 requests per second on the Heroku free tier.

Getting started

Install PostgREST on your platform according to the documentation at postgrest.org, or use the Docker image postgrest/postgrest; the Hackage package is postgrest, and postgrest --help prints invocation help.

How it compares

The facts name no comparable tools or paid products that PostgREST replaces. It stands alone in this registry.

When to use it — and when not to

Choose PostgREST when PostgreSQL is the source of truth and when security, versioning, and validation can live in database roles, schemas, views, and declarative constraints. A self-hoster must operate PostgreSQL and provide JWT-based authentication; the README does not mention SMTP, object storage, or other external services. Skip it when the application needs imperative server-side logic, an Object Relational Mapper, or authorization and integrity rules outside the database, and note that the repository carries 400 open issues.