Worldant is being rebuilt around one idea: a world is a directory that can grow into database schema, callable logic, durable workflow, reactive streams, and generated client contracts.
The project is actively in progress. The current documentation describes the direction we are implementing, not a stable production API.
The goal
Agents should be able to add capability by editing files, running a validated build, and receiving a typed world interface. Humans, agents, generated UI, and services should all call the same commands, mutations, workflows, and subscriptions.
The boundary
Worldant owns the runtime, database truth, durable workflow replay, reactive subscriptions, and npm remote-function protocol. Ragent owns generated UI and consumes Worldant through that protocol.
The proof
The design is proven when an agent can create a small application world, expose typed functions through the live package endpoint, subscribe to data changes, run long workflow jobs durably, and let a generated UI participate through the same committed database truth.