[BUG] Claude Code (claude.ai/code mobile): Timeout on large context — simple commands work, planning documents fail

Resolved 💬 3 comments Opened Mar 24, 2026 by sunpark20 Closed May 23, 2026

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?

What’s Wrong?
Claude Code on mobile web (claude.ai/code) times out when the context becomes large — specifically when working with planning documents or other substantial context. Simple, short commands execute normally in the same session, but any task requiring a large context window (e.g., following a multi-step plan document) becomes impossible to complete.
Initially suspected this was a network issue due to satellite internet (living on a boat), but the pattern clearly indicates a bug: simple commands succeed while complex/context-heavy tasks fail under identical network conditions.

What Should Happen?

Claude Code should process tasks with large context (planning documents, etc.) without timing out, or at minimum provide a clear error message indicating the context size limitation rather than silently timing out.
Actual Behavior
∙ Tasks involving large planning documents or accumulated context always time out
∙ Simple, short commands in the same session execute successfully
∙ This rules out a pure network issue — the problem is specifically tied to context size

Error Messages/Logs

Steps to Reproduce

  1. Open Claude Code on mobile (claude.ai/code)
  2. Load or build up a large context (e.g., a detailed multi-step planning document)
  3. Attempt to execute a task that requires processing the full context
  4. Observe: timeout occurs
  5. In the same session, issue a simple command (e.g., a single-file edit or a short question)
  6. Observe: simple command succeeds

Session Links
https://claude.ai/code/session_016KeMecSCw1pgRcoswFe58T
https://claude.ai/code/session_01Wmtzr3PGvRqWtxwQyVuHjo
Environment
∙ Platform: Mobile web (claude.ai/code)
∙ Network: Satellite internet (boat environment) — but simple commands work fine on the same connection
∙ Claude Code version: Latest (as of March 2026)
Related Issues (searched, not duplicates)
∙ #14407 — Timeout when context remaining < 10% (similar symptom but desktop/Windows-specific, and involves API connection errors after retry). My issue is mobile-specific and the timeout occurs before any retry mechanism.
∙ #29971 — Context bloat from skills/plugins duplication. Related root cause (context overhead) but different trigger — my issue is user-generated context (planning docs), not system overhead duplication.
∙ #28389 — Reduced effective usable context window. Related concept but filed as a feature request about transparency, not a timeout bug.
∙ #36469 — Hangs on large MCP tool responses. Similar “large payload → hang” pattern, but MCP-specific.
∙ #5615 — Bash timeout configuration guide. Not applicable — my issue is about API/streaming timeout on large context, not bash command execution timeout.
∙ #26729 — Streaming resilience feature request for network disconnects. Related to network handling but my issue is specifically context-size-dependent, not random network drops.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

latest 26.3.25

Platform

Anthropic API

Operating System

Other

Terminal/Shell

Other

Additional Information

The key diagnostic signal is that context size correlates with failure, not network speed. On the same satellite connection, small-context tasks succeed while large-context tasks fail. This suggests either:

  1. A server-side timeout threshold that is too aggressive for large context payloads on slower connections
  2. A mobile-specific issue with handling large request/response sizes
  3. The response streaming breaks when the initial payload (with full context) exceeds a certain size over high-latency connections

This makes large planning workflows completely unusable on mobile with slower connections, even though the connection itself is stable enough for simple interactions.

<img width="585" height="1266" alt="Image" src="https://github.com/user-attachments/assets/0ed7a2fd-32a3-404b-bc92-85fa6cdca8d1" />

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗