Features
Eight AI development roles (PM, architect, designer, frontend, backend, QA, security, DevOps)
Uses Claude Code, Codex CLI, or OpenCode as execution backends
112 governance rules enforced at write time, CI, and via MCP server
Configurable quality gate (default 90%) with scorecard
418-document knowledge base with BM25 + optional local semantic RAG (as of v1.0.32, includes documentation count)
Two human-in-the-loop confirmation gates (docs and frontend preview)
Auditable delivery proof pack (ZIP + HTML scorecard) with compliance mapping to SOC 2, ISO 27001, EU AI Act
Memory persistence via facts.jsonl and smart transcript compression
Parallel critic reviews in read-only sessions to avoid reasoning leakage
Self-healing TUI with Windows support (PTY, /exit cleanups)
MCP server mode to expose governance to other tools
Per-step falsifiable evidence contracts (file existence, build success, route response)
Uncertain command detection and three-strike melt breaker
Rollback and /revise commands for iterative refinement
Local-only Rust binary, no cloud or API key needed
@ai_fn decorator for LLM-powered Python functions
@ai_classifier decorator for text classification
Structured data extraction via Pydantic models
Agent loops with tool calling and function calling
Built-in streaming (SSE) support
Async-first API for concurrent applications
Rate limiting and automatic retries
LLM response caching for performance
Concurrency control for workload management
CLI tool for monitoring and debugging
SQLite state store for persistence
OpenAI and Anthropic model support
Embeddings generation for semantic search
Local execution with no cloud dependency
Self-hosted deployable as a library
