TL;DR

Claude Code shows several messages that trace to a 429, with different fixes. 'You've hit your session limit' or 'weekly limit' is your plan allowance, with a reset time in the message. 'Server is temporarily limiting requests' is a short throttle that the docs say is not your quota. 'Request rejected (429)' is a rate limit on an API key or cloud project. Read the exact wording, then use the matching fix.

Claude Code shows several different messages that trace back to a 429, and they have different fixes. "You've hit your session limit" is your plan's usage allowance, and the message gives the reset time. "Server is temporarily limiting requests" is a short throttle that the docs say is unrelated to your plan quota. "Request rejected (429)" is a rate limit on your API key, Amazon Bedrock project or Google Cloud project. Match the exact wording before you do anything. Last verified 14 November 2026 against the official Claude Code error reference, with the npm registry listing 2.1.325 as the latest Claude Code version.

The Claude Code status hub tracks live incidents. This guide covers the limits themselves, using only what the docs state.

Which 429 do you actually have?

The error reference separates these messages, and the fix differs for each:

MessageWhat it isWhat to do
You've hit your session limit · resets 3:45pmPlan usage allowance, shared across all modelsWait for the reset time shown, or buy usage credits
You've hit your weekly limit · resets Mon 12:00amWeekly plan allowance, shared across all modelsWait for the reset time shown, or buy usage credits
You've hit your Opus limit or Sonnet limitAllowance for that model family onlyRun /model and use a model outside the family
Server is temporarily limiting requests (not your usage limit)Short throttle unrelated to your plan quotaWait briefly, then check the status page if it persists
API Error: Request rejected (429)Rate limit on your API key, Bedrock project or Google Cloud projectRun /status, confirm the active credential, check provider limits
API Error: Repeated 529 Overloaded errorsAPI capacity, not a limit on youWait or switch model, see the 529 guide

The times in the first rows are the docs' own examples. Your message shows your own reset time. The same reference also lists spend-limit messages, such as a monthly spend limit, under its usage limits section. They are separate from the rows above.

What does "You've hit your session limit" mean?

The docs say subscription plans include a rolling usage allowance, and that Claude Code blocks further requests until the reset time shown in the message. The session and weekly limits are shared across all models, so switching models does not restore access.

The Opus and Sonnet limits are different. Each applies only to requests to that model family, so switching to a model outside the family with /model keeps you working. The docs add that each model has its own prompt cache, so the next request after a switch re-reads the whole conversation with no cache hits.

What resets, and when?

As far as the docs state, a plan limit resets at the time printed in the error. The error reference does not describe a way to reset it early, and it gives no length for the session window. Anthropic does not document one there, so do not rely on a figure from a forum post. Use the clock in your own message, or run /usage, which the docs say shows your plan limits and when they reset.

The docs list these for when you cannot wait:

  • Run /usage-credits to buy additional usage on Pro, Max and Ultra, or to request it from your admin on Team and Enterprise.
  • Upgrade your plan for higher base limits, which is on claude.com/pricing.
  • For a model-family limit, switch family with /model.

The docs also say that in an interactive session signed in with a claude.ai subscription, Claude Code can wait in the open session and continue the interrupted task shortly after the reset (from v2.1.234). The reference points to the interactive mode page for how to start, cancel or turn off that wait.

A 529 or a short throttle does not count against these allowances. The docs state that a 529 "doesn't count against your quota", and the throttle message says in its own text that it is not your usage limit.

Why might a 429 come from a stray API key?

If you see Request rejected (429) while you expected your subscription to be in use, the docs point at the credential. A stray ANTHROPIC_API_KEY in your environment can route requests through a low-tier key instead of your subscription. Run /status to see which credential is active. The authentication precedence page explains that once you approve it, ANTHROPIC_API_KEY is used ahead of your subscription login.

To fall back to your subscription, the authentication page gives unset ANTHROPIC_API_KEY, then /status to confirm which method is active. That page documents the Unix-style command only. On Windows, use your shell's own way of removing an environment variable. Only remove a key you own and can recreate in the Console. If an API key is the credential you meant to use, the limits are in your provider console, and the docs advise requesting a higher tier there. They also suggest reducing concurrency, for example fewer parallel subagents.

What about unattended and CI sessions?

The docs say Claude Code retries transient failures up to 10 times with exponential backoff before showing an error. For unattended runs, CLAUDE_CODE_RETRY_WATCHDOG=1 retries 429 and 529 capacity errors indefinitely. Two further variables apply to plan limits while the watchdog is set. CLAUDE_CODE_RETRY_WATCHDOG_MAX_USAGE_LIMIT_WAIT_MS caps how long one request waits for a plan limit to reset, and needs v2.1.301 or later. CLAUDE_CODE_RETRY_WATCHDOG_USAGE_LIMIT_RECHECK_MS sets how often Claude Code checks whether the limit has already lifted, takes plain digits from 60000 to 10800000, and needs v2.1.305 or later. These details come from the docs and were not run, because a real plan limit cannot be triggered on demand.

One documented exception: with the watchdog set, a standard-speed request that gets a 429 reporting a spend limit or exhausted usage credits fails at once, including a gateway spend cap that resets on a schedule. The docs say that before v2.1.239 the watchdog retried these indefinitely.

How to tell an outage from a limit

A limit names itself: session, weekly, Opus, Sonnet, or a rate limit on a key. The plan limits come with a reset time, and the key limit points you to a credential to check. An outage shows up as 529, 5xx or a login failure, and as an open incident on status.claude.com. If the message says 529, read the 529 guide. If it says login or 401, read the login loop guide.

More practical writing for people building with AI tools is on the Lead Source blog.

Fixed? Two optional things

Check Anthropic’s status and get an alert when an incident is resolved: Claude Code status and alerts, an independent resource by Lead Source.

Official source: status.claude.com.

Written by the team at Lead Source, where the day job is lead source tracking: working out which marketing actually brings customers.