How is "live streaming" different from the "status" Remote Control used to show?
Status is coarse-grained information — running, completed, failed, and that's about it. You know a task is still going, but not what it's doing right this moment. Live streaming instead sends every tool call a Subagent makes (reading a file, running a command) and every result it gets back, step by step, in real time — meaning what you see on a remote device is the same granularity as what you'd see sitting at the terminal in person.
This distinction matters most in situations that need real-time intervention. If a subagent hits a permission prompt mid-task and you only have status shown, you might not notice it's stuck until you're back at your computer. With live streaming, you can see that prompt directly on your phone and respond to it right there.
Why is this feature limited to foreground subagents — why not extend live streaming to background subagents too?
The key is that foreground and background subagents represent two different user intentions from the start. You send something to the foreground typically because you want to watch it run right now and be ready to step in; sending it to the background is a deliberate choice that says "I don't want to be interrupted by this right now." If background subagents also got live streaming, the system would keep pushing details at you even after you'd explicitly signaled you didn't want to be bothered — a direct conflict with why you backgrounded it in the first place.
There's also a practical volume consideration: most people run several background tasks at once. If each one streamed tool-call details live, the Remote Control side would be flooded with information, which would actually dilute the signal from the foreground task that genuinely needs attention. Limiting live streaming to foreground subagents keeps the feature scoped to a limited number of tasks you're actually paying attention to right now, rather than indiscriminately expanding to every Subagent still running.
How do you actually trigger and see this live streaming? What's the process?
The first prerequisite is running version 2.1.251 or later, since this feature was added in that release. The actual process: you start a Subagent on your local terminal, and it stays in the foreground (not deliberately sent to the background via run_in_background or fork mode), then connect to that session via Remote Control from your phone or another device. Once connected, your remote device's screen shows every tool call and every result from that foreground subagent, updating live, with no extra configuration or toggle needed.
If your session has both foreground and background subagents running at the same time, the remote device will show an asymmetric experience: you'll see step-by-step detail for the foreground one, while the background ones still only show status. That's expected behavior — it doesn't mean the background subagents have a problem or the connection is malfunctioning.
Does this mean you no longer need to stay glued to your computer, and can fully manage Claude Code from your phone?
This feature genuinely narrows the gap between "not at your computer" and "can't see the details," but it doesn't eliminate it entirely — it solves the problem of visibility, not the problem of whether you can perform every operation. You can watch a foreground Subagent's every step live on your phone, and use commands added in the same release like attach, stop, and respawn for basic session management, but that doesn't mean the mobile interface can do everything a local terminal can, especially for scenarios requiring heavy input or complex interaction.
The more realistic way to think about it: this feature gives you a workable option for the middle state of "away from your computer but still wanting to track progress and step in if needed" — it's not meant to replace going back to your computer for full, hands-on operation. If a task genuinely needs intensive back-and-forth adjustment, returning to your computer is still more efficient. If you just want to check progress, respond to a permission prompt, or hit an emergency stop, live streaming on your phone is more than enough.
Among the 71 changes in Claude Code 2.1.251 (released August 28, 2026), one item is easy to lose inside a long changelog but meaningfully changes day-to-day use: Remote Control now streams a foreground Subagent's tool calls and results live to connected devices — background subagents remain unchanged, still showing status only, with no detail streamed out.
Remote Control lets you remotely operate and monitor a running Claude Code session from your phone or another device. Before this update, whether a subagent was running in the foreground or background, all Remote Control could show was high-level status — running, completed, failed. There was no way to see, in real time, exactly which file a subagent was reading, which command it was running, or what result it got back at any given step; for that level of detail, you had to be back at the local terminal.
Starting with 2.1.251, if a subagent is running in the foreground (the kind you'd be directly watching, as opposed to one sent to the background), every one of its tool calls and every result it returns now streams live to whatever client is connected via Remote Control. That means you can be away from your computer and still see, on your phone, exactly which file the subagent is reading right now, which line of a command it's running, and what came back once it finished. Background subagent behavior hasn't changed at all — it still shows status only, with no step-by-step streaming. That distinction is a deliberate design choice, not an oversight in this update.
Foreground and background subagents were never meant to serve the same purpose. A foreground subagent is typically a task you're actively watching and ready to step into; a background subagent is one you deliberately dispatched with the expectation that you won't be watching it constantly. If background subagents also streamed every step live, that would become information overload in a scenario where you likely have several background tasks running at once — the whole point of sending something to the background was not wanting to be interrupted by it. Limiting live streaming to foreground subagents keeps the feature serving one specific use case — you actually want the detail right now, you're just not physically at your computer — rather than indiscriminately pushing every subagent's details at you.
The scenario that benefits most directly is when you've kicked off a foreground subagent at your computer to handle a task, then need to step away — a meeting, getting coffee — but still want to track progress, and might even need to respond to a permission prompt the subagent hits along the way. Previously, stepping away meant losing visibility into the details entirely, with nothing to check until you got back. Now, through a phone or other device connected via Remote Control, you can see exactly what the subagent is doing at that moment, without needing to stay glued to your computer. This update moves in the same direction as the claude attach, logs, stop, respawn, and rm commands added in the same release — all of them gradually decoupling "not at your computer" from "can't see the details or manage the task."