All guides

How to monitor Claude Code from your phone

Check progress, read tool output, and keep a long-running Claude Code session moving from mobile — without constantly SSHing into your machine.

Based on
CLI
0.1.0
Preview ref
f36aa45

The thing that makes Claude Code useful for real work is also the thing that makes it annoying:

Sessions can run for a while.

Sometimes that’s great (it’s chewing through a refactor). Sometimes it’s the worst (it’s stuck, it needs a quick answer, or it’s doing something slightly off‑track and you want to correct it early).

When I’m away from my laptop, I want a lightweight loop from my phone:

Open the session → see what it’s doing → decide if I need to intervene → send one follow‑up.

This is the workflow I ended up with after trying the usual options (SSH, remote desktop, GitHub mobile) and not loving the trade-offs on a phone.

TL;DR

  1. Set up Happier (or attach via Direct sessions), then open the Claude session from your phone.
  2. Use the transcript + tool timeline to see what actually happened.
  3. If the session needs input, send a short follow‑up or handle the approval.
  4. If you need the full terminal view, attach from the owning machine.

The problem (in plain language)

There are three reasons you check a session while it’s running:

  • Progress: “Is it doing the right thing?”
  • Intervention: “It’s drifting — I should correct it now, not after 200 files.”
  • Unblocking: “It’s waiting on my approval / a question.”

On a phone, the failure mode is either:

  • you don’t check at all (and waste time), or
  • you check via SSH and end up half-working in a tiny terminal.

What I mean by “monitor Claude Code from your phone”

From your phone, you can:

  • read progress in the transcript
  • see tool activity (what ran, what changed)
  • handle approvals and user-action requests
  • send follow-ups

The actual work still runs on the machine where the repo and tools live.

What I tried first (and why I changed my approach)

  • Sometimes I did nothing and waited. It’s fine for low-risk tasks — but when it drifts, you only discover that at the end.
  • Sometimes I SSH’d in from my phone. It works — but it turns into “work from phone”, not “monitor from phone”.

What I ended up wanting was a lighter loop: open the session, see what it’s doing, unblock it if needed, send one correction — then get out.

  • Happier: gives me that lightweight loop from my phone (progress + unblock + one follow-up), while Claude Code keeps running on my machine. Trade-off: depends on the owning machine being online/connected.

If you’re new to Happier (what it is, and where to get it)

If you haven’t seen Happier before: it’s an open-source companion app (mobile/web/desktop) for AI coding agents like Claude Code, Codex, and OpenCode.

The workflow here is: use Happier to monitor and send follow-ups from your phone, while Claude Code keeps running on the machine that owns the workspace.

For this guide, that means you can supervise a long-running Claude Code session (and catch drift early) without living in SSH.

Get set up:

A workflow that works (step-by-step)

1) Open the session

From your phone, open the running Claude session.

The first thing I look for is whether it’s actively producing output or blocked.

2) Use the transcript to spot drift early

Read the latest messages like you would read a PR description:

  • what it thinks it’s doing
  • what files it touched
  • whether it’s expanding scope

If you see scope drift, send a short corrective message immediately (it’s cheaper than fixing it later).

3) Check tool activity (what actually happened)

Use the tool timeline to answer:

  • what commands ran
  • what files were written
  • whether it’s looping

Happier normalizes tool events into a consistent vocabulary (like Bash, Edit, Read, Diff) so you can scan what happened without reading raw provider logs.

If the tool views are too noisy on a phone, you can adjust detail level in Settings → Session → Tool Rendering (Title / Summary / Full).

4) Handle approvals and user-action requests

If the session is blocked:

  • approve/deny the permission request
  • or answer the user-action request (AskUserQuestion / Exit Plan Mode)

If you find yourself approving too often for low-stakes operations, consider switching to Safe YOLO for the duration of the task (and review diffs later).

5) If you need full fidelity, attach from terminal

Sometimes you really do need the full terminal experience.

From the owning machine:

happier attach <session-id>

Troubleshooting

Symptom: I’m not getting notifications when the session blocks

Likely causes:

  • notification types aren’t enabled
  • you’re signed into a different Relay/server profile on mobile than the one the session is running on

Fix:

  • enable notifications, then confirm you’re on the correct server/profile on mobile

Symptom: I can open the session but it doesn’t update

Likely cause:

  • the owning machine/daemon is unreachable

Fix:

  • bring the machine back online / restore connectivity; monitoring depends on the execution surface being reachable

Symptom: the tool cards are too verbose on a phone

Likely cause:

  • tool rendering detail level is set to Full

Fix:

  • open Settings → Session → Tool Rendering and switch to Summary (or even Title) while you’re on mobile

FAQ

Is this the same as running Claude Code on my phone?

No.

You can do the monitoring/supervision part from your phone. Claude Code still runs on the machine where your repo and tools live.

Why does my session look “stuck” even though it’s running?

The most common reason is it’s waiting on you:

  • a permission request (allow/deny)
  • a user-action request (AskUserQuestion / Exit Plan Mode)

Open the session and check the latest transcript messages.

mobileclaudemonitoringsessionsworkflows
Last updated: 2026-04-03