[BUG] API Error: Stream idle timeout - partial response received" on Claude Code Web

I've also been getting this on the web interface near constantly for the last 2 days, across Sonnet, Opus and Opus(1M)

Try with "continue" or "resume", this is working for me.
The other "little detail" that seems that works too is to change to Sonnet 4.6.

That's not a fix. And it doesn't work after several continues. And I've been on sonnet4.6 from the start

This is still an issue and has been for me since yesterday, independent of model chosen. It's happening no matter which model I try to use.

Still persisting as of today. No improvement across multiple models or remote environments.

At this point the stream timeout is making Claude Code Web practically unusable for longer tasks. The "continue" workaround only goes so far before the context degrades and the output suffers.

This thread has several duplicates now — clearly a widespread issue. Would really appreciate an official acknowledgment or status update from the Anthropic team. Is this being actively worked on? Is there a known fix in progress?

same issues here. stuck on a task that i urgently need to get done. claude code web interface, MS Edge. Opus 4.7

Something I noticed that I haven't seen mentioned here or in the related threads (#47385, #47252, #46987): when a session dies with Stream idle timeout - partial response received, my prompt itself is missing from the session afterwards. All of Claude's tool calls and partial reasoning from the failed turn are still there, just no user message at the top.

The weird part: sometimes the prompt looks like it's still in the UI right after the failure. But if I hit Ctrl+R, it disappears. So it was only ever in local browser state, never written server-side.

That seems to point at the persistence layer rather than anything output-side. Looks like user messages get committed to durable storage on successful turn completion, so when the turn fails midway, the input gets rolled back along with the output. Which would also explain why continue and resume don't help here, there's nothing on the server to retry against.

Repro is the same as OP. Just check after a failure whether the prompt is actually still there, or hit Ctrl+R and see if it vanishes.

Environment: Claude Code Web, Edge on Windows 11. Happens multiple times per session on small multi-step frontend refactors that are already chunked about as small as I can make them.

EDIT: Another hunch is that it seems to be related to the new UI that has been implemented since Thursday April 23th 2026. This UI is identical to the Claude Desktop UI that was released a few weeks earlier. Back then the desktop timed out far more than the web version (so I moved back to web). But now that that same UI is on the web, the time outs have caught up again. I know that sounds silly, and I have no way to A/B test it anymore because the old web UI has sadly become inaccessible.

Which would be a good feature request. Allowing users to opt in for old versions whenever new UI's are imposed. This is a basic convention. Even Reddit allows people to use their (decade) old UI still, a ghetto way of avoiding any critical issues that come with new UI updates.

Same failure mode:

• Multi-turn planning session on a non-trivial private repo with a

custom CLAUDE.md, context at roughly 33% of the window.

• Asked the assistant to rewrite a long markdown planning document

from scratch as a handoff for a new session.

• Model performed 1–2 exploratory tool calls (file read, scan),

then began the Write call.

• Stream terminated with `API Error: Stream idle timeout - partial

response received` before the write completed in the UI.

• Retried 4 times across 2 separate sessions on the same repository

over 2 consecutive days. All failed at roughly the same point.
Reducing the scope of the write did not help.

• Sessions remained alive, but the same turn could not be

completed — consistent with this being a keep-alive / heartbeat
gap between tool-result-accepted and the next assistant token,
not a model or tool-execution issue.

Workarounds suggested by support (manual chunking of the write,
raising CLAUDE_STREAM_IDLE_TIMEOUT_MS) do not apply to the Cloud
/ browser surface. Confirmation that a fix is in progress for this
specific surface would be appreciated.

Workarounds suggested by support (manual chunking of the write, raising CLAUDE_STREAM_IDLE_TIMEOUT_MS) do not apply to the Cloud / browser surface. Confirmation that a fix is in progress for this specific surface would be appreciated.

Yes, the workaround suggestion is missing the point. Chunking doesn't solve anything. Before April 23rd Opus 4.7 was handling far larger tasks without any snags. After that it times out on far smaller tasks. Tasks that by the way the chat interface can easily handle. It's not that the model is being overwhelmed. It's something that happens in the UI.

> Workarounds suggested by support (manual chunking of the write, raising CLAUDE_STREAM_IDLE_TIMEOUT_MS) do not apply to the Cloud / browser surface. Confirmation that a fix is in progress for this specific surface would be appreciated. Yes, the workaround suggestion is missing the point. Chunking doesn't solve anything. Before April 23rd Opus 4.7 was handling far larger tasks without any snags. After that it times out on far smaller tasks. Tasks that by the way the chat interface can easily handle. It's not that the model is being overwhelmed. It's something that happens in the UI.

Confirming the same regression timing on my side: this started
24 April for me, no issues before that on similar or larger
tasks. The "chat handles it fine" point is also accurate —
I've moved long writes off Claude Code Cloud onto claude.ai
as a workaround in the meantime. Hope the engineering team
picks up the April 23 cutoff as a starting point.

I suspect the fasted solution is to revert the new UI to the old one. Or at least give users the ability to chose between both. Being able to rule out it's the new UI would narrow it down a lot. And if it's the new UI then the problem would be solved right then and there.

Seeing the issue in iOS app. Moving between Claude Code in iOS and web, issue is the same.

Noticing it most frequently but not exclusively when hitting a session limit mid task -> return when new session ticks over and prompt ‘resume’/‘continue’ etc or an adjustment prompt all result in the same message: ‘API Error: Stream idle timeout - partial response received’.

When I’ve hit a session limit I’m never able to recover that instance; sometimes there’s work product diff that can be copy pasted out but typically not. When it occurs mid-session I can sometimes recover it, but not reliably.

Everything works splendidly in the Claude Code extension in VSCode (or any other IDE).

It clearly is a web problem.

Everything works splendidly in the Claude Code extension in VSCode (or any other IDE). It clearly is a web problem.

When I do Terminal or CLI they are fine, so it's indeed the Desktop or Web IDEs from Claude in my humble opinion.

I have also been getting this error for the last couple of hours.

this is getting worse and worse, no one is working on this?