- stabilize semantic workflow profiles beyond Excel benchmark fixtures
- harden OAuth provider persistence and deployment guidance
- expand native-vs-MCP benchmark coverage to more public demo backends
- publish release-tagged benchmark artifacts instead of intermediate run directories
- formal compatibility policy for MCP-facing surfaces
- broader SDK examples for OpenAI-compatible agent loops
- package and deployment recipes for hosted services
- native planning/runtime primitives beyond benchmark-driven semantic workflows
- broader language SDK parity