One chat per agent, and Nox does many things at once
Each agent on nox.cool now has one chat. Ask something quick and it answers right there; ask for real work and it starts a job, a live card in the chat that ends with the result. An agent can work on many tasks at once, each in its own browser tabs, so nothing you ask waits its turn.

Key points
- One chat per agent: quick questions are answered right there, and real work becomes a job, a live card in the chat that ends with the result.
- Ask anything while jobs run. Follow-ups use a finished job's result, so "which flight was cheapest?" just works.
- An agent works on many tasks at once, each in its own browser tabs. Only the whole desktop is taken in turns, and your work goes first.
- A question from work that runs on its own no longer holds anything up: after 20 minutes the work carries on safely, or stops and says what it needs.
- Coding on your Mac: see and continue your Claude Code and Codex sessions from your phone, with nox.cool for Mac 0.1.9.
Each agent on nox.cool now has one chat, and an agent can work on many things at once. Ask something quick and it answers right there. Ask for real work and it starts a job: a live card in the chat that follows the work and ends with the result. Nothing you ask waits its turn behind something else. Everything below is available today in nox.cool's early access.
One chat for each agent
Open an agent from the side menu and you're in its chat. A sum, a fact, a short draft or a question about earlier work gets an answer straight away. Anything that needs the browser, the agent's computer or research becomes a job, and the agent says so in a line.

Jobs that end with the result
Each job's card says what it's doing: working, needs you, or done. When it finishes, the card shows what it found, and your next message can build on it: "which one was cheapest?" or "book that one". Open a card to watch the job at work, or tell the agent "stop the flights job" or "tell the flights job to look at Porto too". Several jobs can run at once.

Many tasks at once
An agent used to work on one task at a time, so a second request said "Waiting its turn". Now each task is busy only with itself:
- New tasks start at once, even while others are running.
- Tasks browse side by side, each in its own tabs on the agent's one computer. Clicking, typing and reading all work in a tab that isn't in front.
- Only the whole desktop takes turns: clicking by position on the screen, or recording it. Your own work goes first, and a task waiting for it says "Waiting for the desktop".
- Typing to a task that asked you something answers it.
- Questions from work that runs on its own (routines, triggers, check-ins) wait 20 minutes. Then the work carries on with the safest choice and says so, or stops and tells you what it needs.
Your coding sessions, from your phone
Turn on Settings › Your Mac › Coding in nox.cool for Mac 0.1.9, then open Coding on nox.cool from any device. You see each Claude Code and Codex session on your Mac and what was said; type what to do next and it continues on your Mac. It won't interrupt a session working in a terminal, and what tools printed is never sent. Add a port such as 5173 and agents can open the app you're building at mac.nox:5173.

Also new
- Previews only you can open: a site or game waiting for your OK no longer opens for anyone with the link.
- A guard on what goes out: card numbers, security codes, passwords and ID numbers are taken out of emails and messages before they're sent; bank and health details always ask.
- Keep watch (Settings › Suggestions): quiet watching of your inbox, calendar and Slack that speaks only when something needs you, plus nudges for emails nobody answered.
- Check it under any reply: each fact, figure and quote checked against its source.
- Where the money went and Where you're signed in, both in Settings.
- Meetings: a summary, decisions and promises from each meeting, with promises to tick off.
- Bring your setup from Grok Bot, MausBot, OpenClaw, Hermes or Claude Code: agents, skills, routines and facts.
- "What would it have done?" on a new rule, from your last 90 days of OK cards.
Try it
Sign in to nox.cool, open Nox and ask something quick, then something big. The full story is in an AI agent that does many things at once.
Questions and answers
How do I talk to an agent on nox.cool now?
Open the agent from the side menu. Its home is one ongoing chat: type a question and it answers there; ask for something that takes real work and it starts a job, shown as a card in the chat. The + next to an agent still starts a separate task if you want one.
Can Nox work on more than one thing at a time?
Yes. Each task is busy only with itself. Two tasks can browse at the same moment, each in its own tabs on the agent's computer, and a quick question is answered while a long task is still working. Only steps that need the whole desktop, like clicking by position or recording the screen, take turns, and your own work goes first.
What happens if a job needs my OK?
Its card in the chat says it needs you, and the OK card waits in the job and in your Inbox, the same as any task. Jobs ask before spending, sending or signing exactly as tasks always have.
Can I continue a Claude Code session from my phone?
Yes, with nox.cool for Mac 0.1.9. In the Mac app, turn on Settings › Your Mac › Coding. Then open Coding on nox.cool from any device: you see each Claude Code and Codex session on your Mac, what was asked and said, and you can type what to do next. It continues on your Mac. What tools printed is never sent.
Where can I try it?
Everything here is available today in nox.cool's early access, which is by invite for now.
Try it with Nox
nox.cool is in early access. Sign in, or ask for an invite on the home page.
Keep reading

What most AI agents still can't do: five things to check
Most AI agents now share the basics. Five things still set them apart, from what you see before it buys to what a job costs the second time.

What should an AI agent show you before it acts?
An AI agent's approval should show the thing itself: the real page, the price, the email. Why a sentence isn't enough, and how nox.cool built it.

What should an AI agent's screen look like?
An AI agent's app has three jobs: take the task, show the work, ask before anything matters. How we rebuilt every screen of nox.cool around those three.