cloudparts

Hasura DDN

Hasura Data Delivery Network: you describe models, commands, relationships and permissions as metadata, and a Rust engine serves them as one GraphQL (or JSON:API) API over databases and code reached through data connectors.

What it is
Web APIs
Type
Commercial product
Made by
Hasura Inc.
License
Proprietary service; the v3 engine and connectors are Apache 2.0
Facts checked
13 October 2026
Website Source code

What it is

A data connector introspects a source, such as PostgreSQL, ClickHouse, DuckDB or your own TypeScript functions, and the ddn CLI turns that into metadata objects in your repo: models that read data, commands that modify it, relationships that join it (even across sources) and permissions that protect it. The CLI builds the metadata into an immutable build, and the v3 engine serves it as a GraphQL API with filtering, sorting, pagination, aggregations and subscriptions, plus a JSON:API interface that uses the same permissions. The engine never runs SQL itself: it plans each query and sends an intermediate representation to the connectors, which speak the open NDC specification. Authentication stays outside: a JWT or a webhook supplies session variables, and the engine checks them against your permissions.

You can run it three ways. Hosted Hasura DDN deploys the API on Hasura’s global infrastructure with autoscaling and its control plane (build versions, schema registry, breaking-change detection, traces). Private DDN puts the data plane in your own VPC or on dedicated infrastructure under an enterprise contract. Or you self-host the open-source v3 engine and connectors with Docker, without the control plane. Pricing is per active model, meaning any model or command called more than 1,000 times a month: Free costs $0 with unlimited models and one developer, Base starts at $5 and Advanced at $30 per active model per month, and hosting third-party or code connectors costs $0.075 per vCPU-hour. In June 2025 Hasura made PromptQL, an AI data-access product, its focus and said DDN and the v2 GraphQL Engine would continue to be hosted and maintained.

Why teams pick it

  • No resolvers to write: introspection generates models and commands, and the engine adds filtering, sorting, pagination and aggregations, nested across relationships
  • Joins across sources: a relationship can link a PostgreSQL table to a ClickHouse table, and one query follows it
  • Permissions and the schema are versioned metadata, so each build is immutable and can be diffed for breaking changes before release
  • The engine (Rust), the NDC spec and the connectors are open source under Apache 2.0, and you can write your own connector with SDKs in Rust, TypeScript, Python or Go

Watch out for

  • A new engine and metadata format: v2 projects need a migration, and v3 was in alpha until 2024 and generally available only from August 2024
  • Self-hosting the engine leaves out the control plane: no built-in CI/CD or build versioning, schema registry or API analytics
  • Per-model pricing can grow quickly with many busy models, and Private DDN needs an enterprise contract
  • Hasura made PromptQL its focus in June 2025; it says DDN will continue to be hosted and maintained, but it is no longer the headline product

Alternatives

What all web apis have in common →Compare side by side →
8baseDormantCommercialA hosted backend on AWS that generates a GraphQL API from the tables you define, backed by Aurora MySQL, with roles, authentication, file storage and custom functions on Lambda.Hasura GraphQL EngineOpen sourceHasura’s open-source v2 engine: point it at Postgres or another supported database and it serves a GraphQL API with queries, mutations and live subscriptions, guarded by role-based row and column permissions.GraphQLOpen sourceA query language and runtime for APIs: the server publishes a typed schema, and each client asks one endpoint for exactly the fields it needs.OpenAPI SpecificationOpen sourceA standard, language-neutral way to describe an HTTP API in one JSON or YAML document, from which tools generate docs, client SDKs, server stubs, tests, gateway configs and AI tools.RESTOpen sourceThe architectural style behind most HTTP APIs: resources with URLs, a few standard methods, status codes and cacheable responses, named by Roy Fielding in 2000.

How it fits with other technologies

From a Postgres database to a GraphQL API with the DDN CLI · bash

ddn auth login
ddn supergraph init my-project && cd my-project

ddn connector init my_pg -i          # interactive: choose the PostgreSQL connector
ddn connector introspect my_pg       # read tables, views and functions
ddn model add my_pg '*'              # models that read data
ddn relationship add my_pg '*'
ddn command add my_pg '*'            # commands that modify data

ddn supergraph build local && ddn run docker-start
ddn console --local                  # GraphiQL and the metadata explorer

ddn project init                     # then deploy to hosted DDN
ddn supergraph build create

See it in a real stack

No architecture uses Hasura DDN yet. Drop it into the playground and wire it up yourself.Open playground →
Where it came from2018–2025

Hasura launched its GraphQL Engine in July 2018 as an open-source product that gave any Postgres database instant realtime GraphQL APIs, with a console for creating tables and running queries. Versions 1 and 2 were written in Haskell, and v2 added data connectors so that query generation for more databases could live outside the engine.

At HasuraCon in June 2023 co-founder and CEO Tanmai Gopal announced the Data Delivery Network: a re-architected engine with zero cold start that could run on an edge network, metadata based on a new Open Data Domain Specification, and a Native Data Connector specification for any source. In December 2023 Hasura open-sourced the v3 engine, now written in Rust, together with the NDC spec and its first connectors. DDN entered beta on 4 April 2024 and became generally available on 20 August 2024, pitched as a metadata-driven data access layer.

On 2 June 2025 Gopal announced PromptQL, an AI platform that writes data-access programs from natural language, as the company’s new focus and the “spiritual successor to GraphQL”. Hasura said the GraphQL Engine and DDN would continue to be hosted and maintained; the v3 engine in the graphql-engine repository was still receiving commits in August 2026.

Why it’s built this way

Rust instead of Haskell
Hasura said the rewrite was about performance, scalability and reliability for cloud-native APIs, and about tapping the wider Rust community.
Connectors outside the engine
In v2 the SQL generation for PostgreSQL, SQL Server and BigQuery was tied to the core engine; with NDC every source is a separate, forkable connector, and the engine only sends it a query plan.
Metadata as the API’s source code
The launch post argued that if Hasura is given resources, methods, relationships and authorization policies as metadata, it can provision the API itself, so governance comes from the same files that define the API.
Price per active model, not per request
The beta post set the goal that APIs should not get more expensive as usage grows, so DDN bills for models that are used, with unlimited requests.

Timeline

  1. 2018

    Hasura GraphQL Engine launches as open source for Postgres

  2. 2023

    Hasura DDN and v3 announced at HasuraCon in June

  3. 2023

    v3 engine open-sourced in Rust, with the NDC spec, in December

  4. 2024

    DDN beta in April; generally available on 20 August

  5. 2025

    Company shifts its focus to PromptQL on 2 June

Sources12 · facts checked 13 October 2026

Improve this page on GitHub