AGENT OS · LOCAL RUNTIME

Click through the product.

Recruiter walkthrough of a local-first agent dashboard. These are real screenshots from the running app — dry-run chat, sandboxed workspace, and machine control with Run command disabled.

View source on GitHub Run it in 3 commands

Mission Control home — Home — prompt-to-workflow plus live local checks for Cursor, Claude, Codex, and Hermes.

Unified Chat — Unified Chat — one box, four agents, dry-run labels. Cursor is honest when chat is not wired.

Workspace sandbox — Workspace — only the Agent OS sandbox. HTML preview uses a sandboxed iframe.

Goals — Goals — Codex Goal Mode shell. Overnight execution is not claimed while the gate is off.

Kanban — Kanban — To Do / Doing / Done from the real local board API.

Memory vault — Memory — counts and search first. Content only after you open an item. Obsidian is not connected.

Studio — Studio — parked / not configured. No fake Midjourney or ElevenLabs.

Machine Control — Machine Control — status checklist. Run command stays disabled.