BioFlow is better as an Apache 2.0 open-core product than as a pure open-source project.
The reason is simple: the local deterministic engine is useful on its own, but the hosted service is where teams pay for coordination, tenancy, billing, and compliance surfaces.
- Local workflow validation, replay, verification, and report generation
- The CLI entrypoint and packaged binary
- The MCP adapter and repo-intelligence surface
- Deterministic sample workflows, fixtures, and tests
- Core schemas, hashing, audit, and artifact handling
- Hosted API and worker processes
- Org provisioning and API-key flows
- Billing and entitlement enforcement
- Team sync, remote run registry, and profile registry
- Enterprise reporting and tenant-scoped operational controls
- It keeps the useful local tooling broad and easy to adopt.
- It preserves a clear upgrade path to hosted collaboration.
- It lets you sell on operational value instead of hiding the core mechanics.
- It keeps the codebase organized around one set of contracts instead of two diverging implementations.
- Keep the MCP layer thin and delegate to the same CLI/service paths the product already uses.
- Treat local workflows and hosted workflows as separate boundaries, not separate products.
- Keep external contributions aligned with
CONTRIBUTING.mdanddocs/CLA.mdso inbound licensing stays clean. - Keep the public README and MCP docs aligned with the same boundary.
- Keep the public core buildable and testable from source.
- Keep the CLA workflow visible in the public repo and out of runtime code.
- Publish client launch examples and auth setup for the MCP server.
- Keep hosted-only capabilities documented in BioFlow-Cloud.