Claude reached its tool-use limit for this turn: the fix
It caps tool calls in one reply, not your 5-hour or weekly usage. Click Continue and Claude resumes where it stopped. Smaller tasks keep it away.
"Claude reached its tool-use limit for this turn" means one single reply ran out of tool calls: searches, file reads, connector requests. It is not your usage limit. Your 5-hour session and weekly allowance are untouched. The fastest fix is to click Continue under the banner, or send a short "continue", and Claude carries on from where it stopped.
This page covers what triggers the banner, the fix, and how to make it rarer. claude/for is independent and not affiliated with Anthropic; sources were checked October 22, 2026.
TL;DR: The banner is a per-reply cap on tool calls, not a quota. Click Continue. If it keeps coming back, split the job into smaller steps and switch connectors to On demand under + > Connectors > Tool access.
What the message looks like
The wording is exactly "Claude reached its tool-use limit for this turn", shown as a banner inside the chat with a Continue button. It appears in claude.ai on the web, in Claude Desktop on Mac and Windows, and in project chats, according to user reports collected in GitHub issue #33969.
You only see it when Claude is using tools: web search, research, Google Drive or Gmail, MCP servers, code execution. A plain text chat with no tools never triggers it. One commenter on that issue also reports the turn stopping silently, with no banner at all.
Tool-use limit vs usage limit
These are two different walls, and the fix for one does nothing for the other.
| Tool-use limit for this turn | Usage limit | |
|---|---|---|
| What it caps | Tool calls inside one reply | Total use over time |
| When it resets | At the next turn | After the 5-hour window or the weekly reset |
| What you see | A banner with Continue | A message that you've hit your limit, with a reset time |
| The fix | Click Continue | Wait, upgrade, or buy extra usage |
Anthropic defines usage limits as caps on "how much you can interact with Claude over a specific time period" in its usage and length limits article. That page does not mention a per-turn tool cap at all. If you are actually out of usage, our guide to Claude's 5-hour and weekly limits is the one you want.
Why you see it
Each tool action Claude takes inside one reply counts as a call. A request like "go through my last 30 emails and summarize each thread" can need dozens of them, and the reply stops when it hits the ceiling.
Anthropic documents the mechanism only for developers. Its stop reasons documentation describes a pause_turn signal, returned "when the server-side sampling loop reaches its iteration limit while running server tools such as web search. The default limit is 10 iterations for each request." The GitHub issue's author links the app banner to this same signal. That link is their reading, not something Anthropic has stated.
The Claude moves worth stealing, by email
Workflows tested on real jobs, sent only when they are worth it. Free.
How to fix it
Click Continue
It sits under the banner. Claude starts a new turn and resumes the task, with the conversation so far still in context.
No button? Type continue
Send a short message such as: Continue from where you stopped. Don't restart the task from scratch.
Check the partial work
If a tool call was cut off mid-way, the last search or file read may be missing. Ask Claude to redo that one step.
Still looping? Split the job
Stop and ask for one batch at a time, as described below.
For developers, Anthropic's documented fix is the same idea in code: "continue the conversation by sending the response back as-is." Add the paused assistant response to your messages and send another request.
How to stop it happening again
You cannot raise the cap in the app. What you can do is ask for less tool work per reply.
- Break big jobs into phases. "Find the five most relevant files" in one message, then "summarize file one and two" in the next. Each message gets its own budget.
- Be specific. "Search our Drive for the Q3 board deck" costs fewer calls than "find everything about Q3".
- Switch connectors to On demand. Anthropic's tool access article puts the setting under the + button, then Connectors, then Tool access. On demand loads only the connectors Claude finds relevant. It is aimed at saving conversation space, so treat any effect on the banner as a side benefit, not a promise.
- Turn off what the chat doesn't need. Anthropic's usage article advises turning off connected apps when a conversation doesn't need them, and asking Claude not to search the web when current information isn't needed. Our connectors guide shows where they live.
Splitting a task this way is also how you get better answers in general, which our guide on how to use Claude covers in more depth.
The honest limits
Anthropic has no help-center article for this banner. Everything about the app's cap, including the "about 20 calls" figure and the March 2026 drop, comes from user reports, not from Anthropic. The phasing advice above is a common workaround, repeated in the GitHub thread and in third-party guides like JD Hodges' write-up, not an official recommendation.
There is also no setting to switch the cap off, and upgrading your plan is not documented to raise it. Most people reporting it in the GitHub thread are on Max plans. Continuing does have a cost: tool use counts toward your usage, so a task that needs five Continues still draws down your 5-hour window.
Questions people ask
Does the tool-use limit count against my usage limit?
The banner itself is not a usage limit, and it does not mean your allowance is spent. The tool calls Claude makes do count toward usage, though, because Anthropic lists tool use among the factors that consume your budget.
Do I lose my work when I click Continue?
It shouldn't. The conversation stays in context and Claude resumes, which matches how Anthropic's API docs describe continuing a paused turn. If a tool call was interrupted when the cap hit, that one step may be missing, so check the last thing Claude was doing.
Will upgrading to Max remove the tool-use limit?
Nothing Anthropic publishes says so. Max 5x and Max 20x subscribers in GitHub issue #33969 report hitting the same banner at roughly the same point. Upgrading buys more usage over time, not more tool calls per reply.
What does "claude tool-use limit" mean?
It is the cap on how many tool actions Claude can take inside one reply before pausing. Searches, file reads and connector requests each count. It resets at your next message, unlike the 5-hour and weekly usage limits.
Anthropic changes limits like this without much notice. The newsletter sends one short note when it does.