Parallel panes
Run several agents on several problems at once, in real terminal sessions in a drag-and-drop grid rather than tabs you have to babysit. Here: Claude shipping a fix while Codex reviews and vitest re-runs on every edit.
DAI is a desktop cockpit that runs Claude Code, Codex and Grok side by side. Coordinated swarms, an agent-driven browser, remote servers over SSH, and one agent that runs it all and proves its results against your tests.
One membership · Code · Voice · Assistant · Creator · Workspace · Business · Hub
The DAI Agent is an operator for the whole cockpit, not a chat box bolted to the side. It answers from your repo, takes on real work, and proves the result with your own tests.
every task passes the scan: dispatched → tested → verified → yours to apply
receipts from one real swarm run, screen-recorded. you can watch the footage instead of taking our word for it.
DAI treats agents like a team you direct. Many hands, one board, one screen.
Run several agents on several problems at once, in real terminal sessions in a drag-and-drop grid rather than tabs you have to babysit. Here: Claude shipping a fix while Codex reviews and vitest re-runs on every edit.
One click spawns a role-primed team (coordinator, builders, reviewer) on a shared board, each agent on its own git branch.
An embedded browser your agents can see, click and screenshot. Point at any element and the exact selector lands in your prompt.
Point a space at any server. Panes spawn there, the file tree follows, and your workflow doesn't change.
Discover servers with a real handshake, toggle them per space, authorize OAuth in-app. Secrets live in your OS keyring.
A local database backs every session, so you can switch spaces all day without anything restarting. Scrollback survives. Local-first, with no cloud backend.
screen-recorded product footage. a fresh film ships with the next release
One membership for every DAI app. You bring your own agent subscriptions. DAI never proxies model traffic.
Prices in USD, excluding VAT. One membership for the entire suite. Every plan is bring-your-own-subscription for the agent CLIs, and DAI never proxies your model traffic. When a trial ends or a plan is cancelled you drop to a free tier that keeps all your work readable. Hub Stripe checkout is parked — the waitlist is the signup path for now.
An operator built into DAI on the Pro plan. One input: it answers from your repo, dispatches real coding tasks with a live activity feed, runs your project's own test suite against every result, and only touches your working tree when you hit Apply. It also drives the app itself (panes, panels, spaces, the browser) and gets better with every task it completes.
No. DAI drives the official CLIs (Claude Code, Codex and Grok) with the subscriptions you already pay for. DAI never proxies your model traffic and never marks up tokens.
On your machine. Workspaces, panes and swarm state are stored in a local database under your user profile; secrets go in the OS keyring. DAI has no cloud backend reading your code.
Claude Code, Codex and Grok CLIs out of the box, plus plain shell panes for anything else. DAI detects the CLIs installed on your machine.
Windows 10/11 today. macOS and Linux are on the roadmap. Remote workspaces already let your panes run on any POSIX server over SSH.
Chat IDEs give you one assistant in one window. DAI orchestrates many real agent sessions in parallel, with swarm coordination, an agent-controllable browser, remote machines, and the DAI Agent operating the cockpit itself. One membership also covers Voice, Assistant, Creator, Workspace, Business and Hub — Bench and OS come later.