Rad

Rad is an experimental relational database that rethinks the traditional stack by combining the database server, migration tooling, and type safe client generation into a single cohesive workflow. You define your schema in a YAML file, run one command to apply migrations, and get generated client code in Go or Rust that queries the database without raw SQL strings. The goal is to remove the friction between data modeling, persistence, and application code so that changes to your schema propagate automatically through the entire stack.

The problem it tackles is the disconnect most developers experience between their database and their application. You write a migration in SQL, then update your ORM models, then fix the type mismatches in your queries, then hope the production migration does not break something a test did not cover. Rad collapses those steps. The schema file is the single source of truth, migrations are generated from it, and the client library is regenerated to match. If you change a column name, the build fails until you fix every call site, which is exactly what you want.

Under the hood, Rad uses a machine friendly intermediate representation instead of raw SQL strings. Queries look like method chains in Go or Rust, compiled to a relational node graph that the planner optimizes. Persistence is backed by SlateDB on object storage rather than a custom disk format, which means you can run stateless compute against durable cloud storage without provisioning NVME volumes. The system supports standard relational operations including joins, CTEs, aggregations, and nested relationship reads that return structured data rather than flat rows.

It is aimed at developers building new projects who want a tighter feedback loop between schema and code, teams tired of maintaining separate migration and ORM tooling, and anyone experimenting with object storage backed databases. If you are already running a production workload on Postgres or MySQL, Rad is not a drop in replacement. It is research stage software designed for exploration, not for data you cannot afford to lose.

What sets it apart is the unified toolchain philosophy. Most databases treat migrations and client generation as afterthoughts or third party add ons. Rad treats them as first class features shipped in the same binary as the server. The generated clients are strongly typed, queries are validated at compile time, and the migration system handles online schema changes with data rewrites built in. There is also an AI agent integration that documents every CLI command so you can automate the workflow with skills compatible tooling.

Access is open source under a permissive license. You install the rad binary, init a project, define your schema, and start the server. The CLI handles migrations, code generation, and administration. Documentation is available on the website, and the GitHub repository includes a demo project. Be aware that this is a proof of concept with no API stability guarantees and zero production workloads. It is ready to try, not ready to trust with important data.