SDCStudio · public beta
Turn CSV data, Markdown templates and plain ideas into semantic data models that keep working across systems, domains and languages.
What it does
AI-assisted
Vertex AI discovers candidate semantic relationships, suggests ontology mappings and drafts documentation. A domain expert still signs off, which is the point: the expert authors the model, the machine does the typing.
Inputs
Import what you have or build from scratch in the visual editor. Thousands of rows process in seconds.
Standards
XML Schema 1.1, RDF, OWL, SHACL and BFO (ISO/IEC 21838-2), plus the rest of the stack. There is no private format anywhere in it.
Outputs
XSD schemas, XML instances, JSON and JSON-LD, HTML documentation, RDF triples, SHACL constraints and GQL statements.
Catalog
Browse and reuse published components at no cost. The more components exist, the cheaper it is for everyone to build the next model. That is the whole economic argument.
Validation
Every model validates against the open-source sdcvalidator, Apache-2.0, on PyPI. Run it locally, offline, with no account.
How it works
1
Bring your data
Drop in a CSV, paste JSON, upload a PDF of requirements, or start from nothing in the visual editor.
2
Draft the model
The AI analyzes your data, proposes relationships and ontology mappings, and drafts an SDC4-compliant model for your expert to correct.
3
Export and deploy
XSD, XML, JSON, RDF, SHACL, GQL or HTML. Deploy anywhere. The model is yours in open standards.
Pricing
Buy credits and spend what you use. Reuse costs nothing, which is deliberate: the catalog only gets cheaper for everyone as it grows.
Browse the catalog
Free
Explore published reference models and reuse existing components at no cost. Download schemas and validate locally with the open-source validator.
Build models
Pay as you go
Mint new components and assemble data models. Reusing catalog components stays free. You pay when AI creates something new, or when you publish a finished model.
Settle in production
$1.00 per Receipt
Issue a Settlement Receipt through the Verifiable Settlement Layer: the instance validated against its published model, the governance decision recorded, the Receipt signed and bound to both parties. 1,000 credits per Receipt. Verifying one is free, with no account.
Security posture
Read from the code and the deployment, kept in a cryptographic inventory, and stated here so a security questionnaire can start from facts rather than from a form.
Transport
TLS 1.2 minimum on the load balancer in front of the service, with hybrid post-quantum key exchange (X25519MLKEM768) negotiated whenever your client offers it, since 9 October 2026. Traffic between the service and its database, cache, storage and key management stays inside Google Cloud on its internal TLS.
Signing
Every Settlement Receipt and practitioner credential carries two issuer signatures over the same digest: ES256 from a P-256 key held in a Cloud KMS hardware security module, and ML-DSA-65 (FIPS 204, post-quantum) from a second Cloud KMS key, software-held because that is the only level Cloud KMS offers for it. Both must verify. Every key ever used stays in the published key document.
Data
The database, cache, uploaded files and secrets use Google-managed encryption at rest. We never see a card number: Stripe collects and holds payment details and we keep a reference. Secrets are injected at deploy from Secret Manager, never stored in an image or a repository. Rendering a page makes no request to a third party; the typefaces ship with the application.
What we do not claim
No customer-managed encryption keys today. No third-party certification; the controls above are stated as built, not as audited. Verification of anything we sign needs no account and no call to us: the verifier is open source and the keys are published.
Browsing the catalog is free and needs no account. Start there.