Our first release lands the walking skeleton: the complete 3S loop, running locally.
What's new
- First-run onboarding gate. Connect and health-check at least one working model before entering the workspace — including a zero-cost local Ollama path.
- Scope → Spec → Skill. Create a project, generate a spec, approve it, and hand off to implementation.
- Local-first task graph. Requirements, specs, tasks, and agent runs are tracked on your machine; keys live in the OS keychain.
- Bring-your-own-agent. Orchestrate Claude Code, Gemini CLI, or a custom agent — result capture reads back from Git.
Getting started
Install the app, pull a local model with ollama pull qwen3:8b, and pick Local Ollama in onboarding. See the getting started guide for the full walkthrough.