A registry, package workflow, and semantic graph for JSON Schema-based contracts.
A registry for JSON Schema-based contracts.
A package workflow for installing and resolving schema dependencies.
A semantic graph of annotations, usage reports, links, provenance, and trust signals.
A shared surface for humans, tools, and agents to reuse the same contract.
Rusl supplies shared contracts and context. Your stack still chooses the runtime machinery.
Not a new schema language.
Not your runtime validator.
Not a serialization format.
Not primarily a code generator.
Use Rusl wherever a data boundary can drift.
Install contracts into a project or resolve them live.
It works like the package managers you already know.
Think Cargo, npm, Bundler, or Mix deps, but for shared data contracts. The manifest names what the project depends on. The version constraints define what is acceptable. The install step makes the resolved contracts available locally.
The product vocabulary is small on purpose.
First-class annotation types turn agent feedback into reusable contract context.
Guidance for what context an agent should load before using a contract.
A report that a contract was used in a project, tool call, generator, validator, or workflow.
Typed explanation of what a field or contract means in a specific domain.
A typed relationship between contracts, fields, annotations, docs, or external sources.
Evidence that helps a consumer decide whether a contract or annotation is dependable.
A structured request for missing meaning when an agent or human cannot safely proceed.
A claim that connects a contract or annotation back to its source of authority.
Version-to-version guidance for compatibility, breaking changes, and upgrade work.
Agents should resolve the shared contract before producing local shape.
This is the behavioral contract Rusl is designed to support.
- 01Query before guessing.
- 02Reuse before inventing.
- 03File context requests when blocked by missing meaning.
- 04Contribute domain interpretations when real usage teaches missing context.