Claude "response incomplete": why it happens and the fix
Claude's reply broke off before it finished, usually a dropped connection or a service hiccup. Copy what you got, refresh, then retry or ask it to continue.
"Response incomplete" means Claude's reply broke off before it finished. The text stopped streaming, usually because the connection dropped or Claude's servers had a hiccup. It is not a usage limit and not a length cap. The fastest fix: copy whatever text did appear, refresh the chat, then retry the message or ask Claude to continue from its last line.
This page covers the wording variants, how this differs from the two pauses that look similar, and how to recover a half-written artifact. claude/for is independent and not affiliated with Anthropic; sources were checked October 22, 2026.
TL;DR: An incomplete response is an interrupted reply, not a limit. Save the partial text, refresh, then retry or type "continue from where you stopped". If it fails across several new chats, check status.claude.com before you keep resending.
What the message looks like
The wording depends on the app and on what broke the reply. These are the versions people report, with where we saw each:
- "Response incomplete": the phrase people search, and the one that spiked during the outage of Sunday, June 21, 2026, when Downdetector logged more than 2,000 reports, per the Asbury Park Press report republished on AOL. That report says "response incomplete claude" was trending on Google, but it does not quote the on-screen text.
- "Claude's response could not be fully generated": quoted in GitHub issue #43642 from Claude Desktop on macOS, alongside "a brief network error notification", and in issue #31073 from a claude.ai user.
- "Claude's response was interrupted.": described in GitHub issue #28787 after switching to another project in Claude Desktop while a reply was still being written.
All three describe the same event: a reply that started and never reached its end. Anthropic's help center documents none of them, so treat the exact punctuation as reported, not official.
Incomplete vs max length vs tool-use limit
Three different things can leave you with an unfinished reply. Only one of them is an error.
| Response incomplete | Max length pause | Tool-use limit | |
|---|---|---|---|
| What happened | The reply was cut off by a failure | One reply hit its length cap | One reply ran out of tool calls |
| What you see | An error, often with the text stopping mid-sentence | "This response paused because Claude reached its max length for a message" | "Claude reached its tool-use limit for this turn" |
| Is it deliberate? | No | Yes | Yes |
| The fix | Refresh, then retry or ask to continue | Click Continue | Click Continue |
If you see a Continue button and a message that names a limit, you are on the wrong page: read our guides to the max length pause or the tool-use limit for this turn. Both are designed stops, and Continue resumes them cleanly.
Why you see it
Claude's replies stream to your screen piece by piece. Anything that breaks that stream partway through leaves you with a fragment. The reports point to four common causes:
- A dropped or unstable connection. Wi-Fi switching networks, a VPN reconnecting, a laptop going to sleep. The Desktop report in issue #43642 pairs the error with a network error notification.
- A service incident. When many people see it at once, it is usually on Anthropic's side, as in the June 21 outage.
- Leaving the chat mid-reply in Claude Desktop. Issue #28787 reports that clicking another project or chat while Claude is writing interrupts and discards the reply.
- Heavy tool work in one turn. The author of issue #43642 reports the error every time a single Desktop turn made three or more calls to a local MCP server. That is one user's reproduction, not a documented rule.
Developers see the same thing in the API. Anthropic's API errors page notes that on a streaming response "an error can occur after the API returns a 200 response", and advises retrying an api_error "with exponential backoff". A reply that stops on purpose looks different: Anthropic's stop reasons guide lists max_tokens and pause_turn for those.
The Claude moves worth stealing, by email
Workflows tested on real jobs, sent only when they are worth it. Free.
How to fix it
Copy what appeared
Select the partial reply, or the partial artifact, and paste it somewhere safe before you retry. A retry can replace it.
Refresh the chat
Reload the page or reopen the conversation. Check whether your message is already there before you send it again, so you don't ask twice.
Retry, or ask to continue
Use the retry control under the reply, or type: Continue exactly from the sentence ending in [last few words]. Don't repeat anything.
Edit your message if it fails again
Edit the prompt to ask for less in one go, such as one section or one file at a time, and resend.
Still failing? New chat, then status
Start a fresh conversation with a short summary and your saved text. If new chats fail too, check status.claude.com.
Retry and continue do different jobs. Retry throws the fragment away and writes the reply again from scratch, which is fine for a short answer. Continue keeps the fragment and asks for the rest, which is better when most of a long answer already arrived. Asking with the last few words quoted stops Claude from restarting at the top.
Editing is the safer route when the same prompt keeps failing. Anthropic's artifacts article says you can "edit an earlier message to create a different version of the chat, with its own artifacts", so trying a smaller request does not wipe out the earlier attempt.
When the artifact text is lost
An artifact (the document or code panel beside the chat) can be cut off the same way, and that hurts more, because it is usually the long thing you actually wanted.
If part of the artifact survived, copy it out first. Then paste it back in a new message: "Here is the document so far. Continue it from the last line, in the same artifact, without rewriting what's there." If nothing survived, ask Claude to rebuild it in sections, one per message. Our Claude Artifacts guide shows how artifacts update in place.
For a very long document, ask for a Word file instead of a chat reply. A file you download is one place the work cannot quietly disappear from mid-stream.
How to make it rarer
- Stay on the chat until Claude finishes, at least in Claude Desktop, given the report in issue #28787.
- Ask for less per reply. A shorter reply has less time to break. One section per message beats a whole report.
- Split tool-heavy jobs. If you use connectors or MCP servers, ask for one lookup batch at a time.
- Steady the connection. Pause a flaky VPN, keep the laptop awake, avoid switching Wi-Fi mid-reply.
- Start fresh for big jobs. Anthropic's usage and length limits article explains that length limits "control how long any single conversation can become". A lighter chat is a lighter load on every reply.
The honest limits
Anthropic does not document "Response incomplete" or its variants anywhere we could find. Its error messages article covers usage limits, chat length and capacity ("Due to unexpected capacity constraints, Claude is unable to respond to your message"), but not interrupted replies. Every cause above comes from user reports and outage coverage, not from Anthropic.
We also cannot tell you whether an interrupted reply counts against your usage. Anthropic does not say. And a refresh is not guaranteed to bring a reply back: sometimes the fragment is all that exists. If several things are failing at once, our Claude not working guide and the couldn't connect to Claude fix cover the wider checks.
Questions people ask
What does "response incomplete" mean on Claude?
It means the reply stopped before Claude finished writing it, usually because the connection dropped or Claude's servers had a problem. It is not a usage or length limit. Copy the partial text, refresh the chat, then retry or ask Claude to continue from its last line.
Is Claude down when it says response incomplete?
Sometimes. If it happens in several new chats in a row, check status.claude.com, because outages like June 21, 2026 produce exactly this error. If it happens once in one long chat, it is more likely your connection or that conversation.
Is "response incomplete" the same as "reached its max length for a message"?
No. The max length message is a deliberate pause with a Continue button that resumes cleanly. "Response incomplete" is a failure: the reply broke off. Continue-style recovery still works, but you type the request yourself and quote the last line Claude wrote.
How do I get back a Claude response that was interrupted?
Refresh the conversation first, because the reply may have finished after all. If it did not, retry the message, or ask Claude to continue from the last sentence it wrote. Save any partial text or artifact before retrying, since a retry can replace it.
If you want a short note when Anthropic changes how Claude handles errors like this, the newsletter sends one when it matters.