Hasura

Hasura DDN is a metadata-driven data access layer. You describe the shape of your data once, and the engine generates the API from that description instead of from code you write by hand. The result is a supergraph that federates many sources into one endpoint, with nested filters, pagination and sorting, plus aggregations out of the box. According to Hasura, the Rust-based compiler and runtime make it faster than the v2 engine, which the company had already called the fastest GraphQL engine.

Data connectors are how DDN reaches each source. Hasura ships official connectors for PostgreSQL, Microsoft SQL Server, CockroachDB, ClickHouse, MongoDB, Snowflake and others, plus code and API connectors for Lambda functions and REST services. They follow the open Native Data Connector (NDC) specification, so you can build your own connector and bring in a data source the official set does not cover. Hosting third-party and code connectors costs $0.075 per vCPU-hour and $0.0075 per GiB-hour.

Every model becomes part of a GraphQL API that supports filters, pagination and sorting, plus aggregations, and the same layer exposes a Relay API for connection-based clients. Nested filters, nested aggregations, nested pagination and nested sorting come standard, which cuts the number of round trips a client needs. The free tier already includes these API features with no limit on requests per month.

Permissions run inside the engine at two levels. Field-level rules control which columns a role can read or write, and entity-level rules control which rows are visible. You define the logic declaratively in metadata, so the table, view or function in front of each API field is gated the same way everywhere. For teams with tight access requirements, one layer for authorization removes the drift that builds up when every service enforces its own rules.

The Advanced plan adds federated collaboration and multi-repo CI/CD, so several teams can work on independent subgraphs and merge them into one supergraph. Breaking-change detection, immutable builds and build URLs let you test a change in isolation before it reaches production. The schema registry and changelog track what changed and when, which matters once more than one team owns the API.

Traces, API performance metrics and model and field usage statistics are part of the console. The free plan keeps traces and metrics for 15 minutes, while paid plans extend retention to 30 days. Model and field usage statistics come with the paid plans and show which parts of the schema are actually used, which helps when you decide what to keep. Autoscaling, zero cold-start and a global edge network with geo routing are included on every plan.

Not everything has to run in Hasura's cloud. You can self-host the v3 GraphQL Engine, run the same tooling locally with Docker, and develop against a supergraph on your machine before deploying. The DDN CLI handles init, connector setup and deployment in one workflow. That matters for teams that need to keep sensitive data on their own infrastructure while still getting the managed tooling.

PromptQL is Hasura's AI layer for working with the supergraph. Instead of hard-coding SQL or GraphQL queries, you ask questions in natural language and PromptQL plans and runs the queries needed to answer them. It's built on the same data access layer, so the permissions and relationships you already defined still apply. For teams adding an assistant or a data app, this turns the supergraph into something an AI system can call directly.