Port42 is a Mac desktop where your team and its AI companions work on the same live work. Bring the Claude Code and Codex seats you already pay for, and stop carrying pieces between windows.
Every AI tool your team adopts opens another window, and every app an AI companion builds opens one more. A dashboard in one tab, a terminal in another, the answer in a chat somewhere else. Your people carry the pieces between them by hand.
On a normal working day, heavy on AI, I counted 244 window switches an hour.
Each new tool is one more place to check, sign in to, and paste into.
What one app knows, the next never sees, unless someone copies it across.
AI runs in terminals and tabs on one person's screen, so the rest of the team finds out afterwards.
Everything your team and its AI companions make lives on one desktop as a port, a live surface you can see, click, and share. A pricing model, a deploy log, a launch review. People and AI companions work on the same port at the same time, and every port shows who is driving it.
Give one AI companion the lead. It splits the work, briefs the others in the port's own chat, and they build in parallel on the same live port. You watch it happen, and step in whenever you want.
The lead reads the port's console and checks each piece before it accepts it. At the end you review the thing itself, running, with everyone's changes in it.
Most multi-agent tools hand each agent an isolated task and send you diffs to review afterwards. Here your AI companions work one live surface together, in a chat you are part of.
Every port keeps its versions. Restore any one of them.
Click the port and you are driving. Your AI companions pick up from your change.
Claude Code and Codex become your AI companions out of the box, and other CLIs and agent frameworks can join too, with integration still deepening. Each uses your team's own Claude or ChatGPT accounts, under the plans and terms you already have. Port42 runs no model and holds no key.
Nothing new to approve. Your team keeps the AI seats it already has.
Port42 never reads a provider credential.
Replace Claude Code with Codex on any seat and nothing else changes.
Your AI companions build what the team needs as ports, on the desktop it already has, and you share the live thing. So there is less to add:
More than 100 things your team could build instead of buy, one line each: the periodic table of software →
Each port shows who is driving it right now, you, a colleague, or an AI companion, and every version records who made it.
If someone changed a port first, a later write is refused and picks up from their change.
Grant access to an AI companion, a person, or any outside process, like a script, a service, or a CI job, through its own bearer token. Revoke one grant and every other grant stays as it is.
Your ports, their chats, and their history stay on your machine. A port reaches someone else only when you invite them to it.
Each CLI talks to its own provider under your team's terms. Port42 is not in that path, and adds no new vendor or data processor for your prompts.
Sharing a port goes through a relay. Start on Port42's hosted relay, with nothing to set up. When sharing has to stay inside your network, run your own. The relay is open source. Run it on your own domain, under your own logging policy, and point your team's Port42 at it in Settings. Sharing then never depends on Port42's servers.
Traffic is encrypted end to end either way. The relay forwards what it cannot read, so whoever runs it sees who connected to whom and when, never the work.
Security status. Sharing in Port42 v1 (invites, the relay, and the browser guest) is new code and has not been independently audited. Connections between machines are encrypted end to end with the Noise protocol, the relay forwards bytes it cannot read, and a guest reaches only the port you invited them to. Those are design properties, not audited guarantees. Until an audit is done, treat a shared port like a screen share: don't share credentials, personal or customer data, or anything covered by regulation. Running your own relay keeps sharing traffic inside your network; it does not replace an audit.
One static binary or a Docker image, with no database. It needs a domain with TLS and WebSockets, nothing else.
The relay keeps its state in memory and stores nothing. What you log around it is your call.
Port42's hosted relay works from day one. Add your relay in Settings and new invites use it, with Port42's relay out of the path.
Everyone runs the same models now, and they get cheaper every month. What sets a team apart is the harness around them.
A thousand years ago the horse collar moved the load off a horse's throat and onto its shoulders, and one horse did the work of a team. Same horse, better harness. Today the throat strap is the window. Your people are the wire between the AI and the work, and the more powerful the AI, the more that costs.
Port42 is the collar. Your AI companions and your people work on the same live surface, so the power reaches the work.
Port42 runs on your computer. Your ports, their chats, and their history stay there. Your AI companions use your own accounts, and you decide who reaches which port. Leave whenever you like, and your work stays with you.
The copying stops. The apps stop being separate places to visit. The window switching falls toward zero, because the work is all in one place.
I clocked myself at 244 an hour. Zero is the target. We are not there yet, and we are building straight at it.