The MVP surface from the sync: a web-first dashboard over the outer loop we already run. We wrap Claude Code as a black box, drive it by prompts, and gate on outcomes (builds / serves / review-passes), never reaching inside its loop. The dashboard foregrounds the orchestration: the stack the AI chose and why, loop status, the decisions ledger, and a link to the live Claude Code session.
Not in scope yet: marketplace, a Codex path, teams / billing. Related: #190 (modular extensions, done), #204 (Open Loop, parked on the sync).
Status: the three substantive children are done. Only #214 remains, which is low-priority and infra-gated (needs a claude.ai subscription + Remote Control + Claude Code v2.1.51+), so it lands when that path is available.
The MVP surface from the sync: a web-first dashboard over the outer loop we already run. We wrap Claude Code as a black box, drive it by prompts, and gate on outcomes (builds / serves / review-passes), never reaching inside its loop. The dashboard foregrounds the orchestration: the stack the AI chose and why, loop status, the decisions ledger, and a link to the live Claude Code session.
Not in scope yet: marketplace, a Codex path, teams / billing. Related: #190 (modular extensions, done), #204 (Open Loop, parked on the sync).
Status: the three substantive children are done. Only #214 remains, which is low-priority and infra-gated (needs a claude.ai subscription + Remote Control + Claude Code v2.1.51+), so it lands when that path is available.