Gel

Gel is an open-source, Postgres-native data platform that replaces the traditional table-and-row relational model with a graph-relational schema built on object types, links, and mixins. Formerly known as EdgeDB, Gel is designed for teams that want Postgres’s durability and ecosystem without the friction of raw SQL—particularly those building GenAI applications, multi-language backends, or apps that require end-to-end type safety. It targets full-stack developers, data engineers, and AI/ML practitioners who need a single, strictly typed data model that works across TypeScript, Python, Go, Rust, .NET, Java, and Dart without an ORM. By layering a modern schema engine, a graph query language (EdgeQL), and integrated Auth/AI capabilities on top of Postgres, Gel aims to be the database layer for the Vercel ecosystem and beyond.

Gel works by compiling a high-level schema—defined with object types, abstract types, mixins, computed fields, access policies, and triggers—into efficient Postgres table layouts. Its query language, EdgeQL, supports hierarchical selects, inserts, and updates with link traversal that eliminates JOINs entirely; queries are composable and can be code-generated into typed client libraries for six languages. The platform includes a network server with both EdgeQL and SQL compilers, a built-in connection pooler and cache, and a zero-config CLI that can spin up a local instance in seconds via `npx gel project init`. For GenAI workloads, Gel offers a vector store with configurable automatic embeddings and an automatic RAG endpoint that integrates with OpenAI, Anthropic, and Mistral, plus SDKs for Python and TypeScript. Integrated auth supports OAuth (multiple providers), email/password, magic links, and passkeys, complete with a built-in login UI. The schema migrations engine is purpose-built from day one, providing interactive, framework-agnostic migrations that can be squashed, branched, and applied across local and cloud deployments.

Gel competes most directly with traditional relational databases (vanilla Postgres, MySQL) and ORM-heavy stacks (Prisma, TypeORM, SQLAlchemy), as well as with graph databases like Neo4j and managed Postgres services like Supabase and Neon. It is also listed alongside ETL and data analysis tools such as Team-GPT, Perplexity AI, Elicit, Sonix, Consensus, Matia, Microsoft Power BI, Tableau, R, and Python—though these are not direct database competitors. Gel differentiates by offering a single, type-safe data model that spans all supported languages, solving the n+1 query problem natively, and providing a built-in graphical UI and CLI for schema management. Its integration with Vercel’s marketplace (for easy billing and preview deployments) positions it as a first-class citizen in the Vercel ecosystem, while its fully open-source licensing (14,000+ GitHub stars) appeals to teams that want to avoid vendor lock-in.

The honest trade-offs: Gel has a steep learning curve for developers accustomed to raw SQL or ORMs, as EdgeQL and the object-type schema require a conceptual shift. Brownfield projects face a full migration process—there is no direct communication with the underlying Postgres instance, so you cannot simply point Gel at an existing database. Server administration and connection tooling can feel immature compared to mature Postgres tooling like pgAdmin or DBeaver, and significant DevOps expertise may be required to tune the Gel server and its connection pooler. Finally, Gel lacks the breadth of ecosystem services and third-party tools that vanilla Postgres enjoys (e.g., pgvector, PostGIS extensions, or managed provider-specific features), and its GenAI features, while promising, are still nascent relative to dedicated vector database solutions like Pinecone or Weaviate.

Fully open source (Apache 2.0 license); cloud deployment available through Vercel with usage-based billing; no paid tiers or proprietary features.