An AI agent framework that builds itself — multi-model ReAct + tiered cache-friendly context + async multi-agent + XML workflows. CLI & WebUI, zero LangChain.
Evidence-bound Harness-of-Harness for verifiable long-horizon agent engineering: receipt-bound acceptance, independent QA, durable program state, recovery, explicit limitations. Pre-release v0.3.0a2 (stable: v0.1.0), developed under its own orchestration since 2026-09-07, with product-run merges since 2026-09-25. Research preview.
A CLI tool for scaffolding new projects from Jinja2 template directories. Define typed variables directly in your templates, load default values from YAML data files, and fill in the rest interactively through a terminal form.
