Drop a model in. It tracks the board.

What "drop-in" looks like

• Open the workbench. The model tray is already populated.
• Drag the card you want onto an open lane, or onto the chat dock.
• Watch the dotted handoff line draw itself between the card and the Board.
• A tooltip lights up: "now tracking PR #42 · last commit 14 min ago."

What you skip

• No API key files to manage. OAuth handles auth.
• No .env juggling between projects.
• No per-tool README to read before the first prompt.
• No config file with eighteen optional flags you have to learn.
• No manual board wiring — the card finds the lane on its own.

What it tracks the second it lands

• The open Board cards and which lane is yours.
• The current PR in flight, with head SHA and review status.
• The last commit on main and how far the lane has drifted.
• The open lane assignments (who's on what, who's idle).
• The owner-pending list — anything waiting on your call.

The tools you get day one

• Ollama bundled — local Llama / Qwen / Mistral. Yours from the second the hub boots, offline by default.
• Whisper voice — on-device dictation. Runs on the bundled Ollama brain, no extra wiring.
• Telegram bot — push notifications and reply-to-approve, set up in the install wizard.
• MCP servers from the catalog (fs, shell, search, browser, more) — install with one click via Wizard.
• Remotion render queue — video jobs run while you sleep.

↳ Claude, Codex, and Pi integrate with RHOBEAR — bring your own OAuth account when you want their horsepower. We never resell them.