Direct real coding agents from one calm place. Brainstorm an idea and hand it to a builder, run several agents in parallel, and pick it all up on your phone. Built to ship — not to babysit a terminal.
Bring your own agent
From the first idea to a shipped change — across every agent, model and machine you use.
Shape an idea with Claude — no files touched. When the spec is right, hit Build and a coding agent picks it up in the right folder.
A 1 / 2 / 4 grid or a fully dockable workspace — drag chat, terminal and preview into the layout that fits you, and watch several agents at once.
SSH terminals that survive over tmux — run agents against the infra you already have.
Install as an app over your own secure network. Same chats, context and settings — check in from anywhere.
Theme every color, dock every panel, add skills — then export and share your setup.
Browse a catalog of verified skills from official sources and install them per agent — Claude Code, Codex, Gemini and more. Every skill is previewed (code + security flags) before it touches your machine.
Pick a skin or fine-tune every color token, drag panels into your ideal layout, set your shortcuts. Then Copy your theme as JSON and share it. A calm surface you’ll want to live in all day.
A real workspace for AI builders — without becoming an IDE you have to wrestle with.
Drive Claude Code, Codex, Gemini, Hermes or Pi from one UI.
Hand a subtask to another model — a new chat spins up beside it.
A 1 / 2 / 4 grid or a fully dockable workspace.
Think it through, then hand the spec to a builder.
A clean three-tier sidebar with drag-and-drop.
Each folder gets its own instructions, env and defaults.
Commit, push, open PRs, isolated worktrees, merge to base.
Review every change, then one-click an AI review.
A file tree + viewer with optional inline edits.
A local embeddings index helps agents find the right code.
Natural language → command, plus one-click Fix & Explain.
SSH terminals that survive over tmux on your infra.
Install as an app — same chats, context and settings.
Color skins + a per-token editor you can export and share.
Drag every panel into the layout that fits you.
Install verified skills per agent, in one click.
Connect tools and data via the Model Context Protocol.
Run tests on finish; save commands to the palette.
Reusable prompts and quick file/folder references.
Dictate with on-device Whisper — no key, private.
Start a chat straight from an assigned issue.
See spend and how full the model’s context is.
Talk an idea through with Claude. Shape it into a clear, buildable spec — nothing is changed yet.
Hit Build. A coding agent opens in the right folder and starts working. You review the diffs.
Commit, push, open a PR — and keep an eye on it from your phone.
What sets it apart from an IDE bolt-on or a single-agent chat box.
Brainstorm with Claude, hit Build, and a coding agent picks it up in the right folder. No other cockpit makes that handoff seamless.
SSH terminals over tmux against your servers, and a phone app that mirrors your machine over your own secure network.
Any agent harness, any model, your own keys or subscription. Dices orchestrates — it never locks you in.
Theme every color, dock every panel, add skills, share your setup. Made to live in all day.
A calm cockpit for directing agents — not an IDE you have to wrestle with.
Voice and codebase search run on-device. Your code and keys stay on your machine.
Free during the alpha. Download the desktop app and connect with your own Anthropic API key — or bring your own Codex, Gemini or Cursor CLI.
Download links go live with the first release.