Current Limitations
This page reflects current implementation status.
Implemented
- CLI with
run,mcp,config - engine REST API
- provider adapters (Anthropic/OpenAI/Gemini)
- Docker-based execution
- generated pytest validation
- weighted scoring and retry loop
- runtime MCP and skills discovery
Not Yet Implemented
- gRPC transport between CLI and engine (HTTP used today)
- comparison view and MCP explorer pages in the web UI
- dedicated scorers for infrastructure/research/automation
- persistent run history commands in CLI
- hosted community stack registry
Practical Constraints
- depends on external APIs (provider, registries, optional Tavily)
- registry data can vary over time
- generated tests and completeness scoring depend on LLM quality
Roadmap Link
See explicit implementation matrix: