Pricing
Claude Code "API Error: 529 Overloaded": What It Means and What to Do
Seeing 'API Error: 529 overloaded_error' in Claude Code? Why it isn't your usage limit, how to check status, switch models with /model, and when to wait.
8 min read
Error meaning and fixes checked Oct 30, 2026 in Claude Code’s error reference and Anthropic’s API error docs. The raw error string is quoted from Claude Code GitHub issue #35704 (verified Oct 29, 2026). Team plan details checked Oct 29, 2026 on claude.com/pricing. Every source is listed at the bottom. No affiliate links on this page (how we work).
The exact error: API Error: 529 {"type":"error","error":{"type":"overloaded_error","message":"Overloaded"}}
If your terminal just printed this:
API Error: 529 {"type":"error","error":{"type":"overloaded_error","message":"Overloaded"}}(that’s the raw form users posted in GitHub issue #35704, verified Oct 29, 2026), or the newer, friendlier version from Claude Code’s own docs:
API Error: Repeated 529 Overloaded errors. The API is at capacity — this is usually temporary. Try again in a moment. If it persists, check https://status.claude.com.Here’s the one-line answer (as of Oct 30, 2026): Anthropic’s servers are busy, not your account. Anthropic’s API docs define 529 overloaded_error as “The API is temporarily overloaded” and add that it “can occur when the API experiences high traffic across all users.” Claude Code’s error reference is even blunter: “A 529 is not your usage limit and doesn’t count against your quota.”
So put the credit card down. Upgrading won’t fix this one, and neither will logging out and back in. Everyone on that model is stuck in the same queue. It’s the same feeling as the ticket printer in The Bear when the kitchen is already slammed: it keeps spitting tickets, the line can’t cook any faster, and none of it is your fault.
529 vs 429 vs usage limit: which one did you hit?
Claude Code can stop you for three very different reasons, and they need three different fixes. Here’s how Anthropic documents them:
| Error | Means | Your move |
|---|---|---|
| 529 overloaded | API busy for everyone | Wait or /model |
| Repeated 529 | Retries ran out | Status, retry |
| 429 rate limit | Your key or org’s cap | Check /status |
| Temporary limiting | Short server throttle | Wait briefly |
| Session limit hit | Your plan’s allowance | Wait for reset |
Sources: Anthropic API errors and Claude Code error reference, both verified Oct 30, 2026.
A few details worth knowing:
- 429
rate_limit_errormeans “Your organization has hit a rate limit, reached its usage tier’s monthly spend cap, or reached a spend limit on the Claude Code workspace” (API docs). In Claude Code it shows up asAPI Error: Request rejected (429), and it’s tied to your API key, Bedrock project or Google Cloud project. API Error: Server is temporarily limiting requests (not your usage limit)is Anthropic’s short-lived throttle. The message literally tells you it isn’t your quota, which is nice of it.- Plan limits look completely different:
You've hit your session limit · resets 3:45pm, or the weekly, Opus or Sonnet versions. Those are your plan. For how they work, read Claude Code usage limits.
If your message has “529” or “Overloaded” in it, you’re in the first two rows. Everything below is about those.
What to do right now
Claude Code has already tried before it showed you anything. Its docs say it “retries transient failures up to 10 times with exponential backoff before showing you an error”, and that includes overloaded responses that arrive before any of the reply has streamed. While it’s retrying you’ll see a Retrying in Ns · attempt x/y countdown, and on recent versions (v2.1.198 and later) a 529 also tells you to check status.claude.com. By the time the red text lands, the polite retries are done.
Here’s the documented order of operations (Claude Code error reference, verified Oct 30, 2026):
- Check status.claude.com. If there’s a capacity notice, you have your answer. If you’re on Bedrock, Google Cloud or a gateway, the message names that provider’s status page instead.
- Wait a few minutes and try again. Your original prompt is still in the conversation. For server errors, Anthropic’s docs suggest typing
try againinstead of pasting a long prompt a second time. - Switch models with
/model. This is the most useful fix, because Anthropic says “capacity is tracked per model.” When one model is under heavy load, Claude Code may tell you itself, with a line likeOpus is experiencing high load, please use /model to switch to Sonnet. In the Desktop app’s Code tab, the same message readsOpus is experiencing high load. Switch to Sonnet.and you use the model picker. - If the error came mid-answer, you may see
API Error: Server error mid-response. The response above may be incomplete.Claude Code keeps whatever finished. Replycontinueand it picks up from the last completed block.
What not to do: don’t spin up a second account, don’t hunt for “unlimited” proxies, and don’t paste your session into a random gateway. Capacity is the bottleneck, so none of that gets you a faster kitchen.
Running Claude Code unattended? Tune the retries
If a 529 is killing your CI job or overnight script, Claude Code has documented environment variables for exactly this (error reference, verified Oct 30, 2026):
CLAUDE_CODE_RETRY_WATCHDOG=1: Anthropic suggests it “in unattended sessions such as CI jobs to retry429and529capacity errors indefinitely” instead of failing after the normal retry count. It still fails right away on a 429 that reports a spend limit or exhausted usage credits.CLAUDE_CODE_OVERLOADED_RETRY_BASE_DELAY_MS: starting backoff for 529 retries. Default 500, up to 32000. Needs v2.1.292 or later.CLAUDE_CODE_OVERLOADED_RETRY_MAX_DELAY_MS: ceiling on that backoff. Default 32000, up to 64000. Needs v2.1.296 or later.CLAUDE_CODE_MAX_RETRIES: default 10. Lower it if you’d rather a script fail fast.
The two delay settings have no effect when the watchdog is on, or for requests sent in fast mode. Mid-response recovery needs v2.1.199 or later. And a quick gripe: an “indefinitely” switch that comes with a separate setting to cap the wait tells you everything about how long some of these capacity crunches have run.
Does a 529 count against my Claude usage?
No. Claude Code’s error reference says it directly: “A 529 is not your usage limit and doesn’t count against your quota” (verified Oct 30, 2026). A request the API refused for capacity isn’t eating your session or weekly allowance.
What does cost you is the next request. If you switch models with /model, Anthropic notes in the same docs that each model has its own prompt cache, so the first request after a switch re-reads the conversation with no cache hits. Not a reason to avoid switching, just a reason not to switch back and forth every five minutes like Roman Roy picking a side in the succession fight.
If it happens every day
Daily 529s are a peak-hours problem more than a plan problem, but there are two plan-level facts worth knowing:
- Max and Team list priority access; Pro doesn’t. The claude.com/pricing grid shows “Priority access at high traffic times” as No on Pro and Yes on Max and Team, and Anthropic’s Team page lists “Priority access during high-traffic periods” for Standard seats (verified Oct 29, 2026). Anthropic doesn’t say how much priority that buys, so treat it as a tiebreaker in a crowded kitchen, not a reserved table. Team starts at two seats.
- It’s still not a fix for a real outage. When status.claude.com shows a capacity incident, priority or not,
/modeland waiting are the moves. If you’re weighing the upgrade for other reasons (mostly more usage), see Claude Pro vs Max.
Also check the boring stuff once: run /status and confirm Claude Code is using the credential you think it is. A stray ANTHROPIC_API_KEY in your shell can route you through a key instead of your subscription, and that changes which limits and errors you see. For costs on each route, see Claude Code pricing and Is Claude Code free?.
If capacity crunches keep wrecking your workday, it’s fair to keep a second tool warm. Our Claude Code alternatives list and AI coding tools compared page cover the options, with dated prices.
Here’s what I’d do
Read the message once. “529” or “Overloaded” means wait, check status, or /model to another model, in that order. “Session limit” or “weekly limit” means it’s your plan, and the fix is the reset time, usage credits or an upgrade. “429” means a key or org cap, so run /status first. Most of the panic around this error is people treating a server traffic jam like a billing problem.
A different error, “Invalid API key · Please run /login”, is a sign-in problem, not overload: see Claude Code’s invalid API key error.
If the error says Unable to connect to API (ECONNRESET) instead, it’s usually a network problem on your side; see the Claude Code ECONNRESET fix.
Not an overload but a refused edit? See Claude Code “File has been unexpectedly modified” for why the Edit tool asks for a fresh read.
Seeing “OAuth session expired and could not be refreshed” instead? The Claude Code OAuth session expired fix walks through the login side.
Seeing a 400 rather than a 529? See Claude Code API Error 400.
Timing out rather than overloaded? Claude Code: the operation timed out covers the documented timeout settings.
Seeing a 403 rather than a 529? See Claude Code API Error 403.
Overloaded (529) and rate-limited (429) are different codes: Claude Code API Error 429.
FAQ
What does Claude Code API error 529 overloaded mean? It means Anthropic’s API is temporarily at capacity across all users; it’s a server-side issue, not your account or plan.
Is Claude Code down? Check status.claude.com for current incidents; a 529 means the API is overloaded, which is often temporary and model-specific.
Does 529 overloaded use up my Claude usage? No, Claude Code’s docs say a 529 is not your usage limit and doesn’t count against your quota.
How do I fix overloaded_error in Claude Code? Wait a few minutes and retry, check status.claude.com, or run /model to switch to a different model since capacity is tracked per model.
What’s the difference between API Error: 529 {"type":"error","error":{"type":"overloaded_error","message":"Overloaded"}} and a 429?
A 529 is Anthropic’s servers being busy for everyone; a 429 is a rate limit or spend cap on your key, project or organization.
Will upgrading to Max stop 529 errors? Not guaranteed; Anthropic’s pricing grid lists priority access at high-traffic times on Max and Team but not Pro, and doesn’t say how much priority that means.
Sources
- Claude Code error reference (Repeated 529 wording, “doesn’t count against your quota”, /model fix, retries, environment variables, 429 and usage-limit messages): https://code.claude.com/docs/en/errors (verified 2026-10-30)
- Claude Code troubleshooting (routing for 5xx, 529 and 429 errors, /doctor, /feedback): https://code.claude.com/docs/en/troubleshooting (verified 2026-10-30)
- Anthropic API errors (529 overloaded_error, 429 rate_limit_error, 500 api_error definitions): https://docs.claude.com/en/api/errors (verified 2026-10-30)
- Claude Code GitHub issue #35704 (raw 529 overloaded_error string): https://github.com/anthropics/claude-code/issues/35704 (verified 2026-10-29)
- Anthropic status page: https://status.claude.com (linked, no live status claimed)
- Claude pricing (priority access at high traffic times, Pro vs Max vs Team grid): https://claude.com/pricing (verified 2026-10-29)
- What is the Team plan (priority access during high-traffic periods, two-seat minimum): https://support.claude.com/en/articles/9266767-what-is-the-team-plan (verified 2026-10-29)