Your agent has the schema.It still doesn't know what it means.

Rusl turns JSON Schemas into living contracts: versioned, annotated, reviewable, and resolvable by humans, tools, and agents.

JSON Schema gives agents the shape, not the shared meaning. That shape is cheap to generate; the cost begins when every downstream system has to preserve the intent behind it. Rusl attaches the interpretations, examples, provenance, risk, migration guidance, and feedback agents need before they generate or modify code.

The pain

Same schema. Different assumptions.

Private data shapes create a tax on every integration. AI-assisted development feels fast because the first shape is cheap, then brittle because every downstream surface has to preserve the private assumption.

User-visible or internal?
One agent renders a field in the UI. Another treats it as operational metadata. Both saw the same schema.
Required today or migration artifact?
A field is present, but its history is missing. The agent cannot tell whether to preserve it, migrate it, or hide it.
Fallback or hallucination?
Ambiguity turns into plausible payloads. The code compiles, the validator passes, and the domain meaning is still wrong.
Compounding Token Cost

Compounding Token Cost is the downstream stack you now own.

The token cost to generate a schema is a rounding error. The expensive path is creating a local contract, then becoming responsible for every product and engineering surface that now has to support it: storage, HTTP exposure, UI, validation, mappings, tests, docs, and enough agent context to keep the next pass from drifting or reinventing it.

Storage model and migrations
HTTP routes, docs, and generated clients
Form inputs, display components, and empty states
Validators, imports, exports, and analytics
Adapters for near-duplicate field names
Agent context to preserve the decision later
Collaborative contracts

GitHub made code reuse normal. Rusl does the same for living data contracts.

Code became reusable when it had shared places to be published, discovered, reviewed, discussed, versioned, forked, and depended on. Data contracts need the same collaborative infrastructure.

Rusl applies that move to schemas and their meaning. The schema is the shape. The shared contract layer carries proposals, annotations, bundles, usage reports, source attestations, and trust signals that agents and humans can resolve together.

Publish
Give a contract a stable home under a person or organization account.
Review
Move contract changes through proposals before they become new versions.
Annotate
Attach typed interpretations, examples, policy, trust signals, and feedback without forking the schema.
Reuse
Resolve the same contract from apps, agents, tools, and bundles instead of cloning local variants.
Where drift happens

Drift happens at every boundary that touches data.

Storage validates shape. Domain code assigns meaning. APIs publish contracts. UI turns fields into interaction. MCP tools expose actions to agents. If those boundaries resolve intent privately, they drift separately.

Consumption modes

Use contracts locally or resolve them live.

Both modes support the same goal: stop treating schema meaning as private memory inside one repo or one prompt.

npm-style
Project-local install
Resolve schemas into the project with a CLI and bundle manifest, then use them during build, generation, validation, and review.
live
Runtime resolution
Fetch schema contents directly when an app or agent needs the latest resolvable contract and surrounding semantic context.
Shared dependency

Rusl makes the contract a shared dependency.

Version the shape, attach typed meaning, package related schemas, record feedback, and let agents resolve the same contract before they generate or modify code.

Shared contract layer
The contract is the shared dependency. These are the parts that make it reusable.
Version the shape
Attach typed meaning
Package related schemas
Review contract changes
Record feedback from use
Why MCP makes this urgent

MCP tools expose schemas. Rusl answers what those schemas mean.

Tool parameters and results already have shape. The next failure is interpretation: how the schema should be used, which examples are trustworthy, what risk exists, and what an agent should do when meaning is ambiguous.