Remote Screen lets you interact with another Mac's desktop or bring an individual remote app into your local desktop. It is a human-control tool: you click and type in applications running on the other machine. That is different from sending instructions to a coding agent.
The current Agent.Space Remote Screen page specifies Apple silicon Macs running macOS 14 or later. iPhone and iPad support is marked coming soon. This guide follows that documented scope, not a claim of compatibility with every operating system or network.
Remote Screen is not the only way in. The Agent.Space Mac app also includes remote control for Macs connected to a Workspace: you can open your own connected Macs without a device code, and Workspace members can control a Mac only while its owner shares it with full access. The next two sections cover the standalone app; the section on the Mac app covers the built-in route.
Prepare the two Macs
Install the official Remote Screen app on both supported Macs. On the Mac that will be controlled, enable the option that allows other Macs to control it and review the requested screen-recording and accessibility permissions.
These are consequential permissions. Screen access can expose documents, messages, and account details; control permission allows input into applications. Close unrelated private windows and decide whether this machine is appropriate for remote access before approving the prompts.
Workspace roles do not grant screen control by themselves. In the Mac app route, a member can control another person's Mac only after its owner turns on Allow members to use full access for that computer. The same setting lets members give their Agents full access on that Mac: no approval prompts, any command, and files outside the shared folder. Review workspace roles and permissions before sharing a Mac with colleagues.
Connect to the intended machine
The documented flow discovers Macs on the same network. Across networks, use the controlled Mac's device code. Verify the destination on both sides before approving the first connection; do not rely only on a familiar-looking display name.
Some restrictive networks may prevent the direct connection. Do not fix that by disabling a corporate firewall or broadly exposing the Mac to the internet. Ask the network owner to evaluate the requirement, or use another approved access method.
Control a Workspace computer from the Mac app
If a Mac is already connected to a Workspace through the Agent.Space Mac app, you do not need a device code. On the Mac to be controlled, open the app's Remote tab and allow Screen Recording and Accessibility. Until both are allowed, other devices show it as not turned on.
On the controlling Mac, connect the Mac app to the same Workspace. Then choose the computer under Control Another Computer, or select Remote control for it on the Workspace's Use local computer page. Your own connected Macs appear there; another member's Mac appears only while its owner shares it with full access. Controllers reach the Mac through Agent.Space rather than local-network discovery.
Unlike a first Remote Screen connection, nobody at the controlled Mac is asked to approve a Workspace connection. The Mac's Remote tab lists who is controlling it now and can disconnect them. To stop members from connecting again, turn off full access on the Use local computer page.
Choose an entire desktop or one app
Use the full desktop when a task spans several applications or you need to understand the machine's state. Use a single remote app when the job is narrower and you want it alongside local windows.
The product describes individual app windows, Dock integration, and text/image clipboard exchange. Treat the clipboard as another data path: before pasting, check both its content and the destination. A single-app view should not be mistaken for a security sandbox that limits everything the remote user could access.
For example, a developer might keep a local task brief open beside a remote editor. The practical benefit is less context switching—not proof of faster agent execution. Your model still runs through its configured provider, and the program still executes on the chosen machine.
Verify with harmless actions first
Before opening a production project, use a non-sensitive text document. Confirm that you can identify the remote machine, type a short test sentence, see the result, and disconnect. Try clipboard transfer only with disposable text, not a password or token.
Record whether the connection works on the networks you actually use. The product's performance figures have specific test conditions and should not be treated as a guarantee for your connection. We have not measured your network latency or bandwidth.
Disconnect and revoke access when finished
The product provides visible remote-control indication, disconnection, and revocation of trusted devices. Review trusted devices after lending, replacing, or losing a client computer. Ending a viewing session and revoking future trust are distinct actions.
When helping another person with Remote Screen, have them approve access themselves and agree on what you will open. Do not silently enable permanent access for convenience. In a Workspace, the owner's full-access setting is the approval, so turn it on only for members you would let operate that Mac, and turn it off when they no longer need it.
Use it as the human part of an agent workflow
An agent can investigate code, while a human may still need to inspect an application or make a decision. Keep that distinction explicit in the task: define which actions the agent may take and which require you.
Read how Agent.Space keeps a project together, then review Remote Screen and its current download requirements. If you need a separate machine as well as remote access, evaluate Dedicated Mac; the screen tool and the machine solve different parts of the workflow.