[BUG] Remote Control: AskUserQuestion selections made on mobile not received by CLI — causes infinite wait
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
When using Remote Control from a mobile device (claude.ai/code), selecting options in an AskUserQuestion tool call does not propagate back to the CLI session running on PC.
Specific scenario:
- Claude Code CLI on PC invokes
AskUserQuestionwith multiple-choice options - The question renders correctly on the mobile Remote Control interface
- User selects an option and submits on mobile
- The CLI on PC never receives the selection — the AskUserQuestion prompt remains displayed, waiting for input indefinitely
- The session is effectively stuck with no way to proceed other than manually intervening on the PC (pressing Escape / rejecting the tool call)
This appears to be a state synchronization issue between the Remote Control web client and the CLI process — the mobile client believes the answer was submitted, but the CLI never acknowledges it.
Note: Regular text messages sent via Remote Control work fine. The issue is specifically with interactive tool prompts (AskUserQuestion selections / permission approvals).
What Should Happen?
When the user selects an option in AskUserQuestion via Remote Control on mobile, the selection should be delivered to the CLI session and the conversation should proceed normally — identical to selecting the option directly in the terminal.
Error Messages/Logs
No error messages — the CLI simply remains in a waiting state showing the AskUserQuestion prompt.
Steps to Reproduce
- Start a Claude Code CLI session on PC (e.g., via terminal or VS Code)
- Connect to the session via Remote Control on a mobile device (claude.ai/code)
- Trigger a workflow that invokes
AskUserQuestion(e.g., a skill/agent that asks user preferences with multiple-choice options) - On mobile, select one or more options and submit
- Observe: Mobile shows the selection was submitted, but the PC CLI still displays the unanswered question prompt
- The session is stuck in an infinite wait state
Claude Model
Opus
Is this a regression?
Not sure — Remote Control is relatively new
Last Working Version
_No response_
Claude Code Version
2.1.56 (Claude Code)
Platform
Anthropic API (Max)
Operating System
Windows 11 + WSL2 (Ubuntu) — CLI side
Mobile (Android / iOS) — Remote Control side
Terminal/Shell
Bash running inside tmux on WSL2 (Windows host)
Additional Information
- Environment detail: Claude Code is running inside a tmux session on WSL2 (Windows 11 host). The tmux multiplexer may be relevant to the state synchronization issue — the CLI receives Remote Control text messages fine, but interactive prompts (AskUserQuestion) fail to receive responses submitted from mobile.
- The issue is specific to interactive tool prompts (AskUserQuestion). Normal text input via Remote Control works correctly.
- Permission approval prompts (e.g., tool execution approvals) may also be affected, but AskUserQuestion is the confirmed case.
- Workaround: Reject the tool call on the PC side manually, or type the response directly in the PC terminal.
20 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
This is not a duplicate of the suggested issues. Here's why:
| Suggested duplicate | What it reports | How this issue differs |
|---|---|---|
| #28488 | Permission mode toggles (plan/ask) not syncing from Desktop to CLI | Our issue is about AskUserQuestion tool responses not reaching CLI |
| #28427 | Permission prompts not rendering on iOS | Our issue: prompts render correctly on mobile, selections are submitted, but CLI never receives the response |
| #28330 | Plan mode acceptance not showing on mobile | About plan approval UI, not AskUserQuestion tool |
This issue is specifically about:
AskUserQuestiontool renders correctly on mobile Remote ControlThe root cause appears to be a response delivery failure for interactive tool prompts (not a rendering issue), which is distinct from all three suggested duplicates.
Environment: WSL2 + tmux on Windows 11, connected via Remote Control from mobile.
Related Issues — Interactive Prompt Sync Failures
After investigating remote-control issues, I found several reports that share the same root cause pattern: mobile ↔ CLI state synchronization failure for interactive prompts.
Directly related (same symptom family)
| Issue | Symptom | Overlap with this issue |
|-------|---------|------------------------|
| #28427 | Tool permission prompts intermittently missing on iOS | Interactive prompt delivery failure (theirs is rendering; ours is response delivery) |
| #28330 | Plan mode acceptance not shown on mobile | Mobile → CLI interactive response not propagated |
| #28926 | Mobile stuck in infinite loading loop while CLI is idle | State desync — mobile believes session is active, CLI disagrees |
| #28571 | Messages silently lost after connection drop, no resync | Mobile UI shows success, CLI never receives — identical pattern to this issue |
Common pattern
All five issues (including this one) share:
This suggests a systematic issue in the bidirectional channel for tool-level interactions (AskUserQuestion, permission approvals, plan acceptance), distinct from the text message channel which works reliably.
I experienced a similar issue on v2.1.63, but with a slightly different behavior — instead of an infinite wait, the questions were silently skipped without being displayed.
The CLI output showed:
...with no actual question content shown, and the flow just continued as if the question had been answered.
This was on the CLI side (not the mobile Remote Control interface). Disabling Remote Control resolved the issue and questions appeared correctly again.
Seems like the same underlying state synchronization problem manifesting differently.
Same issue here, latest claude code 2.1.78.
When using my mobile as remote i just see the claude symbol turning and nothing happens.
When i check my session instance on my debian pc i see claude code is waiting for the user to answer some questions.
These questions are not shown on my mobile.
When i answer the questions on the server, it continues on the remote too.
In 2.1.81 (macOS)...
It says like Using AskUserQuestion... This time, options were displayed on the CLI side, so I was able to select them, but on the remote side, it's completely frozen.
I am seeing this issue with Android 1.260330.27 and claude code 2.1.81 on macOS.
No tmux, no WSL, no docker containers, just running the regular version installed by Homebrew.
I'm getting same error with android app version 26040910 and latest claude code version 2.1.113 (latest) running on Windows 10 and Windows 11. I've always had this issue
Adding data points from a separate repro (filed as #50463, closing as duplicate of this):
AskUserQuestioncase you confirmed is one instance of a broader pattern. When the mobile "Code" controller is used to approve a Bash tool permission prompt, the local CLI TUI wedges: the in-flight tool call renders an indefinite spinner (blinking dot), Esc / Ctrl-C / input are no-ops, and the assistant-side context receivesTool result missing due to internal error. The mobile controller continues to function — new prompts/responses flow through it — proving the backend session is alive, only the local view is desynced.claudeprocess and relaunch withclaude --resume.Environment:
Suggests the event routing for interactive tool prompts (permission approvals, AskUserQuestion, probably anything that blocks the CLI waiting on user input) isn't propagating the "resolved from remote" ACK back to the local TUI's render loop.
Another data point from Windows 11 Pro (10.0.26200). Same behaviour — approving a permission prompt from the mobile app dismisses the dialog in the terminal but the session hangs. The approved command never runs. No error, no retry. Have to manually intervene on the PC.
Happens consistently with Bash tool permission prompts. Filed #55833 before finding this thread.
This has been open since February with 5+ duplicates across Windows, Linux, and WSL. The bug makes remote approval completely unusable — any permission prompt approved from mobile hangs the terminal session. That's a pretty core workflow for anyone using Claude Code away from their desk.
Is there any timeline on this, or a known workaround beyond pre-approving everything?
This looks like a different failure mode from a missing mobile UI. The prompt can render, but the interactive response ACK does not make it back to the local CLI/TUI state, so the session gets stuck even though normal text messages may still work.
For an official fix, I would treat AskUserQuestion / permission approval / plan acceptance as one class of remote interactive prompt resolution, and make the local TUI observe an explicit "resolved remotely" state transition. Without that ACK path, mobile can appear successful while the live CLI process is still waiting.
For workarounds, I would not rely only on pre-approving everything. The more robust pattern is to keep an independent control path to the same live process, usually tmux/SSH, so you can still answer, interrupt, or recover the session when Remote Control's prompt channel desyncs.
I built a small open-source tool around that narrower workaround: https://github.com/Snailflyer/faryo
It keeps Claude Code/Codex/shell inside tmux as the live session layer, and uses a phone/desktop browser workbench to review output, send short input, approve/interrupt, and hand off context into the same session. It is not a fix for this Claude Remote Control bug and not a replacement for the official mobile client; it is a self-hosted escape hatch for people who want a separate control surface over the same tmux-backed process.
The fix probably wants one resolved-remotely ACK path for all interactive halts:
AskUserQuestion, permission approvals, plan acceptance, etc.Right now mobile can claim success while the CLI still thinks the gate is open. A typed receipt like
prompt_resolved_remotely { prompt_id, responder, selected_option_ids, resolved_at }would make the unblock auditable and give the local TUI one state transition to observe instead of bespoke flows per prompt type.Agree. A typed receipt is the shape I would expect too.
The important part is that the local TUI transitions from "waiting for local user input" to "resolved remotely" through an explicit prompt id, not by inferring success from the mobile controller path. That also gives maintainers something auditable: if the ACK is missing or rejected, the UI can expose a recoverable desync state instead of leaving the user with an indefinite spinner.
Happening on win 11. Android app. Claude cli and desktop version appear to both experience it. Latest versions as of today's date
Got the same issue since month.
I do not get why they do not fix it.
For me that is a curial part to have not to sit in front of my PC or using a Cloud session
Reproduced:
Suggested Fable5 Prompt for dev team:
"Pls fix remote control, thx"
If Claude asks you something mid-session (mobile, desktop, or web) and you're not there to answer immediately, it just keeps re-asking. So you end up with the same question stacked three times, with no indication of which one is live. Only the last one registers, so unless it's a simple multiple-choice tap, you get to retype your answer over and over. I've also watched Claude shrug and go "since you're not answering, I'll just pick one" entirely unprompted.
This reproduces in basically every session I run, and has for months. How does this keep happening?
Does nobody at Anthropic actually use their own product?
That is something I asked myself too. At least not these features or only in auto mode