[BUG] Dispatch on iOS Claude app stuck in 400 whitespace error loop — no self-service clear (continuation of #45217)
Summary
Dispatch (the mobile-to-PC task-bridge feature in the Claude iOS app) is stuck in a permanent 400 error loop. Every send to the Dispatch thread returns:
API Error: 400 — content contains white-space
The Dispatch thread cannot be cleared, deleted, or reset from any client-side UI (iOS app, claude.ai web, or reinstall). The conversation is server-side persistent, so users cannot recover from this state without Anthropic backend intervention.
Why this is filed as a separate issue from #41992
This was originally reported as #45217, which was auto-closed by the duplicate bot as a duplicate of #41992 and is now locked.
#41992 is about Claude Code CLI session JSONL corruption — the user-facing remediation path is to delete the local .jsonl session file. That works because CLI sessions are client-stored.
Dispatch is server-persisted, single-thread, no client-side state:
- There is no local file to delete.
- The iOS app has no "delete thread" / "clear history" / "new Dispatch" UI.
- claude.ai web has no Dispatch management UI.
- Reinstalling the iOS app does not reset state.
So while the underlying root cause may be shared (empty/whitespace content block in stored conversation history), the user-facing fix for #41992 does not solve Dispatch. This issue tracks the Dispatch-specific recoverability gap.
Steps to reproduce
Unclear what action put the Dispatch thread into this state — it occurred mid-session. Once in this state, it is fully reproducible:
- Open Claude iOS app.
- Send any message into the Dispatch thread.
- Receive
400 content contains white-space. - Repeat indefinitely.
Expected behavior
Either:
- The whitespace-content block should not be persisted server-side in the first place (root-cause fix, shared with #41992), AND/OR
- The user should have a self-service way to clear or reset their Dispatch thread (e.g., a "Reset Dispatch" button in iOS settings or claude.ai).
Actual behavior
Permanent unrecoverable lock-out from the Dispatch feature without contacting Anthropic support.
Environment
- Claude iOS app (Dispatch feature)
- Account-level state (server-persisted)
References
- #45217 — original report, auto-closed and locked
- #41992 — listed duplicate parent (CLI session corruption; same root cause, different recoverability story)
Ask
- Treat the Dispatch user-facing recoverability gap as in-scope (separate from the CLI fix in #41992).
- Provide a self-service clear/reset mechanism for Dispatch.
- In the meantime, document the support escalation path for affected users (current Anthropic support routes Dispatch tickets to the Connector/OAuth team, which doesn't address the issue).
This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗