Wireless computing: the computer stays, the wire goes

The useful part of wireless computing is not a smaller screen. It is keeping the computer that does the work, then reaching it from wherever you are.

Most “work from anywhere” advice assumes the work should move. Copy the project to a laptop. Push it to a branch. Rebuild it on a server. Open the same files on three machines and hope the versions agree.

That is a lot of movement for a simple goal: you want to keep working when you stand up. A different arrangement is simpler. Leave the project, the agent and the model on the computer that already has them. Give every screen an authorized way in.

Your computer is already the server

A development setup is full of long-running work. A package manager is installing. A test suite is running. An agent is waiting on a tool call. An app server is on a port. Moving all of that to a phone is not realistic, and it is not necessary.

Lyre treats the desktop as the host. The project files stay there. The agent process stays there. The local model stays there. Your phone, another computer or the browser client opens a session to that host and shows you what it is doing.

That is the whole trick. The screen changes. The computer does not.

What travels, and what stays

It helps to separate the two, because the privacy story lives in the difference.

Stays on the host: source files, build output, terminals, running processes, provider keys, local model weights and the agent's working context.
Travels to the client: the conversation, the diff you asked to see, the live app preview, and the input you send back.
Depends on your network: how those views reach the client. On the same Wi-Fi, a paired device can connect directly. Away from home, the session can use an end-to-end encrypted relay.
Depends on your provider: the prompts and project context you send to a cloud model. A local model keeps that request on your hardware; a remote provider handles it under its own terms.

The last point matters. “The project is local” does not mean every byte of AI work is local. It means the project and the runtime are. What you send to a model is a separate choice, and Lyre keeps that choice visible.

Three ways in

The desktop app is the full workspace. The phone app is for the work that benefits from being away from the desk: reading a reply, approving a step, opening the live app, checking a diff. The browser client covers a computer where you cannot install the app. Each one connects to the same host and the same conversation.

You can start with the desktop and add devices as you need them. Pairing is explicit, and developer access to a project is a separate grant from being paired. An app-only invitation is narrower still: it opens one running app without opening your source, agents, terminals or desktop. The pairing guide and security guide describe those boundaries.

Build, review, ship, extend

The phone is not only for reading status. You can start an agent, queue a follow-up, dictate a request, attach a screenshot and review the resulting diff. You can open the live app and try the change with your thumb instead of imagining how it will feel.

Publishing works from the same workspace. Open Publish, see what is missing, review the destination and confirm. The host brings the build tools; a phone does not need to become a build machine. The publishing guide lists what each destination actually requires.

Extensibility travels too. The Marketplace and the plugin system belong to the same workspace, so adding a tool does not mean going back to a different app.

Remote desktop is the last mile

Some tools never became web apps. A game engine, a simulator, a CAD program, a design tool with a plugin you depend on. Those are the moments when an app preview is not enough.

Lyre's remote desktop shows the host's whole screen beside the conversation. It is built in, so there is no RDP server or gateway to install. Viewing and control are separate grants, and each session asks for consent. It is currently labelled early access; the remote desktop guide is the honest place to check the current platform support.

The limits are part of the design

A host-based setup has requirements. The computer must be awake, online and running Lyre. A phone cannot start a project on a host it cannot reach. If the connection drops, the work may still be running, but you need to reconnect and read the current state before sending another request.

Availability also varies by platform. Desktop downloads, phone store availability and the browser client each have their own support notes. Remote desktop is early access. These are not footnotes to hide; they are the reason the setup is understandable.

Try the small version first

You do not need a rack of computers to see whether this fits. Start with one project on one computer. Pair one phone. Open the same conversation and the same app from both. Change something on the desktop and watch it on the phone. Then close the phone and confirm the work kept running.

If that feels useful, add a second computer or a local model. The point is not to own more devices. It is to stop treating the computer in front of you as the only place your work can exist.