A voxel, a decision, a resource status: each one is a real, documented schema, not an internal implementation detail. If you want to integrate against the platform, read the contract -- you don't need a partnership agreement to see the shape of your own data.

The open surface is deliberately narrow and Apache 2.0 -- the same playbook dbt and Supabase run (dbt Core's engine, Supabase's self-hostable server). It holds what a record is: its shape, keys and vocabularies, and the building blocks that read and write it. Commercial value lives in the work done on the record, not in the record itself.

The core never forks for a demo -- every surface conforms to the same published schema whether you're running the open building blocks or the commercial platform. Checking, comparing, predicting and proving records, multi-tenant operations and the hosted service are what Ectropy sells; the record shapes and the core building blocks are what the industry gets to own outright.

ectropy-ai/core holds TypeScript building blocks: IFC parsing, model federation, clash detection, row-level tenant security, an AI provider abstraction, an MCP client and the SDK. The records they read and write are the open standard at schemas.ectropy.ai, so integration happens over the documented record, not a code dependency on Ectropy itself.

• GitHub -- ectropy-ai/core
• Schemas -- schemas.ectropy.ai