EdgeDB (Gel) — The Next-Generation Graph-Relational Database

Introduction

EdgeDB (branded as Gel in 2025) is built by two core Python contributors and aims to fix what decades of ORMs and SQL have avoided: the object-relational mismatch. It runs on top of Postgres but exposes an object-oriented schema (types, links, computed properties) and a purpose-built query language (EdgeQL) that makes relational work feel natural.

With over 14,000 GitHub stars, EdgeDB is attracting attention from developers tired of writing ORM models and manually joining rows. Native clients exist for TypeScript, Python, Go, and Rust.

What EdgeDB Does

Schemas are object types with properties and links (typed edges to other objects). The compiler turns these into a Postgres schema under the hood. EdgeQL queries are composable (return shaped JSON directly), type-safe (client codegen), and support computed properties, polymorphism, and access policies.

Architecture Overview

[EdgeDB Server]
  Query compiler: EdgeQL -> SQL
  Migration engine
  Built-in connection pool
  Authentication + access policies
        |
  [PostgreSQL]  (embedded, managed by EdgeDB)
        |
  [EdgeQL Features]
   shaped results (nested JSON out of the box)
   implicit joins via links
   computed properties
   constraints + indexes
   access policies (row-level security, natively)
        |
  [Client Bindings]
   TypeScript (typed query builder),
   Python, Go, Rust, Elixir

Self-Hosting & Configuration

// TypeScript client with codegen
import { createClient } from "edgedb";
import e from "./dbschema/edgeql-js"; // generated query builder

const client = createClient();

const query = e.select(e.User, (u) => ({
  name: true,
  email: true,
  posts: (p) => ({
    title: true,
    created: true,
    filter: e.op(p.created, ">", e.cast(e.datetime, "2026-01-01T00:00:00Z")),
    order_by: p.created,
  }),
  filter_single: { email: "alice@example.com" },
}));

const result = await query.run(client);
// result is fully typed — no manual DTOs, no runtime guessing

# Access policy: users can only read their own posts
type Post {
  required title: str;
  required author: User;
  access policy author_only
    allow select
    using (.author = global current_user);
}

Key Features

• EdgeQL — composable, typed, shape-aware query language
• Links (not FKs) — typed relationships, zero JOIN boilerplate
• Schema migrations — version controlled, auto-generated
• Access policies — row-level security in schema
• Typed clients — TypeScript codegen returns typed results
• Postgres inside — mature storage engine; Postgres features available via escape hatches
• HTTP + GraphQL protocols — optional, talk EdgeDB from anywhere
• Gel rebrand — same project, new name + sharper branding in 2025

Comparison with Similar Tools

FAQ

Q: EdgeDB vs Prisma? A: Prisma is an ORM over your chosen database. EdgeDB is a database with a native typed query language. EdgeDB avoids the ORM-SQL impedance mismatch entirely; Prisma works with any Postgres/MySQL.

Q: Why the rebrand to Gel? A: The team renamed the product to "Gel" in 2025 for a sharper brand identity. The GitHub repo is geldata/gel (formerly edgedb/edgedb). Core functionality is unchanged.

Q: Is EdgeDB production-ready? A: Yes. Multiple startups ship EdgeDB in production. For conservative stacks, vanilla Postgres remains the safer default — but EdgeDB has proven mature for greenfield TypeScript/Python teams.

Q: Can I use vanilla Postgres features? A: Yes via escape hatches. EdgeDB stores data in Postgres, so you can run SQL directly for advanced needs (at the cost of bypassing EdgeDB safety checks).

Sources

• GitHub: https://github.com/geldata/gel (formerly edgedb/edgedb)
• Docs: https://geldata.com
• Company: Gel (formerly EdgeDB Inc.)
• License: Apache-2.0