What to do when Claude Code is down: 9 ways to keep coding in 2026

Disclosure: Enjoy is our product. We checked every way below against its maker’s own pages on November 11, 2026, and we say where another way is the better pick, including the free ones that don’t need Enjoy.

The ways to keep coding, compared

When Claude Code stops answering, start with the message. API Error: Repeated 529 Overloaded errors and API Error: 500 Internal server error mean the trouble is on Anthropic’s side, not yours, and Claude Code’s error reference says of the first: “A 529 is not your usage limit and doesn’t count against your quota.” By the time you see either one, “Claude Code has already made whatever retries apply to that failure.”

Most incidents are short and narrow. In the 30 days to November 10, 2026, Anthropic’s status page lists 11 incidents that named Claude Code. Nine named particular models, such as “Elevated errors for Claude Sonnet 5.5,” and two were about the Gmail and Google Calendar connectors. Eight were marked resolved within an hour of being posted. The longest that named a model, on October 25, took an hour and 39 minutes from first post to resolved, and Anthropic put the impact itself at 30 minutes. So the first three ways below cost nothing and are usually enough. The rest are for the long ones, and for work that can’t wait.

Way	Cost, checked November 11, 2026	How soon you’re back	Best for
Check the status page, and wait	Free	When Anthropic recovers, most often within the hour	A short incident
Switch models with /model	Included in your Claude plan	Right away, if the other model is healthy	An incident on one model
A fallback model chain	Included in your Claude plan	By itself, for that turn	Long and unattended runs
Claude on Amazon Bedrock, Google Cloud or Microsoft Foundry	Pay per token, billed by AWS, Google or Microsoft	Right away, once it’s set up	People who already have a cloud account
OpenRouter	Each provider’s list price per token, plus 5.5% when you buy credits	Right away, once it’s set up	Keeping the same Claude model
Codex, on a ChatGPT plan	Free; Go $8 a month; Plus $20; Pro from $100	Right away, in a new or imported chat	A second lab’s agent
A local model, through Ollama	Free, on your own computer	Once it’s downloaded	Work that can’t depend on anyone’s servers
T3 Code	Free	Right away, in the same thread	Switching agents in one thread, free and open source
Enjoy	A monthly subscription, free for 7 days, with your phone and one computer (pricing)	Right away, in the same conversation	Handing a conversation to another lab’s agent

Prices are each company’s US list prices, before tax. Waiting, switching models, a fallback chain, Codex on ChatGPT’s Free plan, a local model and T3 Code cost nothing extra; the others add to what you pay for Claude. Claude’s pricing has Pro at $20 a month, or $17 billed yearly, Max at $100 and Ultra at $200.

First, read the message

Not every stop is an outage. Claude Code’s error reference sorts them, and the fix differs for each:

What Claude Code shows	What it means, in Anthropic’s words	What to do
API Error: Repeated 529 Overloaded errors	“The API is temporarily at capacity across all users.”	Wait a few minutes, or switch models
Opus is experiencing high load, please use /model to switch to Sonnet	One model is “under particularly high load”	Run /model
API Error: 500 Internal server error	“an unexpected failure inside the API. It is not caused by your prompt, settings, or account.”	Check the status page, wait a minute, send again
API Error: Server is temporarily limiting requests (not your usage limit)	“a short-lived throttle that is unrelated to your plan quota”	Wait briefly and try again
You've hit your session limit or You've hit your weekly limit	Your plan’s usage ran out, which isn’t an outage	See our usage-limit guide
Unable to connect to API. Check your internet connection	These “usually originate in your local network, proxy, or firewall”	Run curl -I https://api.anthropic.com

Anthropic’s status page lists its parts separately: claude.ai, the Claude Console, the Claude API, Claude Code, Claude Cowork and more. So it can tell you whether the trouble is one model, signing in, or everything. An incident can also change shape as it goes. On September 29, Anthropic wrote that “The errors that began at 14:00 UTC were mitigated at 14:36 UTC,” but “A second issue is preventing many users from signing in” and from starting “Claude Code or Cowork sessions.” Its advice that day is worth remembering: “If you’re signed in, please don’t sign out.” It put the end of the impact at 14:59 UTC.

Check the status page, and wait

Best for: a short incident, which most of them are.

Claude Code is already trying for you: “Claude Code retries transient failures up to 10 times with exponential backoff before showing you an error.” While it does, the spinner shows “a Retrying in Ns · attempt x/y countdown,” and for a 529 the line below it “also names where to check service status.” status.claude.com posts each incident with updates until it’s resolved, and has Atom and RSS feeds.

After an error, you don’t need to paste your prompt again: “Your original message is still in the conversation, so for a long prompt you can type try again instead of pasting the whole thing.” For a run nobody is watching, there’s CLAUDE_CODE_RETRY_WATCHDOG: “Set to 1 in unattended sessions such as CI jobs to retry 429 and 529 capacity errors indefinitely instead of failing after CLAUDE_CODE_MAX_RETRIES attempts.”

Price: free. A 529 “doesn’t count against your quota.”

Watch for: the status page can trail the problem. The November 8 incident was posted at 08:28 UTC, and Anthropic later put the start of its impact at 08:18. A clean page with errors on your screen can mean you’re early, or that the trouble is your own network. More retries only go so far: CLAUDE_CODE_MAX_RETRIES is “Capped at 15” unless the watchdog is on.

Switch models

Best for: an incident on one model, which is the usual kind.

One of Anthropic’s own fixes for a 529 is to “Run /model and switch to a different model to keep working, since capacity is tracked per model.” Claude Code suggests it when one model is busy, with a line such as Opus is experiencing high load, please use /model to switch to Sonnet, and in the Claude desktop app, “you switch models with the app’s model picker.” The status page’s titles tell you which model to leave: of the 11 incidents in the last 30 days, nine named the models affected.

Price: included in your Claude plan.

Watch for: some incidents hit several models at once. On October 23, one named “Claude Fable 5.1, Sonnet 5.5, and Haiku 5.5.” A switch also costs a little usage: “Each model has its own prompt cache, so the next request re-reads the whole conversation with no cache hits.”

Set a fallback model chain

Best for: long runs and scripts, where nobody is there to type /model.

Claude Code can make the switch itself. Its model configuration docs say: “When the primary model is overloaded, unavailable, or returns another non-retryable server error, Claude Code can switch to a fallback model instead of failing the request.” Start it with a list:

claude --fallback-model sonnet,haiku

To keep the chain across sessions, set fallbackModel in your settings as an array. “Configure one or more fallback models and Claude Code tries them in order, showing a notice when it switches,” and “Claude Code also applies the chain to subagents.”

Price: included in your Claude plan.

Watch for: “The switch lasts for the current turn only, so your next message tries the primary model first again,” and “Claude Code caps chains at three models.” You can’t confirm it in advance: “Claude Code doesn’t confirm the chain at startup and /status doesn’t display it.” And it changes the model, nothing else. The docs list “Authentication, billing, rate-limit, request-size, and transport errors” among the failures that “never trigger a switch,” so it won’t help when signing in is what’s broken.

Claude on Amazon Bedrock, Google Cloud or Microsoft Foundry

Best for: people who already have an AWS, Google Cloud or Azure account, and need Claude’s models in particular.

Claude Code doesn’t have to reach Claude through Anthropic’s API. It can sign in to Amazon Bedrock, Google Cloud’s Agent Platform, formerly Vertex AI, or Microsoft Foundry instead. The error reference treats each as its own service: server errors come from “Anthropic’s service on the Anthropic API, and the service behind that provider’s endpoint on Amazon Bedrock, Google Cloud’s Agent Platform, Microsoft Foundry, or a custom gateway,” and on those, an error message names “that provider’s service status” in place of Anthropic’s.

For Bedrock, “Run claude. At the login prompt, select 3rd-party platform, then Amazon Bedrock,” or run /setup-bedrock if you’re already signed in. On the AWS side, you select an Anthropic model in the Bedrock console and “submit the use case form. Access is granted immediately after submission.”

Price: pay per token, not your Claude plan. Anthropic’s deployment overview lists the billing as “PAYG through AWS,” “PAYG through GCP” and “PAYG through Azure.”

Watch for: this is set-up work for a calm day. Bedrock needs “An AWS account with Amazon Bedrock access enabled” and “Appropriate IAM permissions,” and the wizard “saves the result to the env block of your user settings file,” so the choice outlasts the session. “The WebSearch tool is not available on Amazon Bedrock,” and “Microsoft Foundry has no interactive setup wizard.” A cloud’s endpoint can have a bad day of its own, which is why Claude Code points you to that provider’s status and not Anthropic’s.

OpenRouter

Best for: keeping the same Claude model when Anthropic’s own API is the part that’s failing.

OpenRouter routes one account to models from many providers and bills per token. Its post on 529 errors makes the case: “Anthropic’s API isn’t the only place Claude runs. On OpenRouter, Claude Sonnet 5.5 is served by five providers, and so are Opus 5.5 and Haiku 5.5.” It lists Anthropic, Amazon Bedrock, Google Vertex AI, Azure and Claude Platform on AWS, and its default routing “deprioritizes any provider that’s had significant outages in the last 30 seconds and retries a failed request at the next provider.”

To point Claude Code at it by hand, OpenRouter gives three environment variables:

export ANTHROPIC_BASE_URL="https://openrouter.ai/api"
export ANTHROPIC_AUTH_TOKEN="$OPENROUTER_API_KEY"
export ANTHROPIC_API_KEY="" # must be empty

Its Ori Harness does the same without touching your usual setup: “Only sessions you start with ori claude run on OpenRouter, and plain claude works as before.”

Price: OpenRouter’s pricing says “Inference is billed at the provider’s list price on every plan,” and that its fee “is charged when you buy credits, 5.5% on Standard and 8% on Business, never on individual requests.” Anthropic’s list price for Opus 5.5 is $4 per million input tokens and $20 per million output tokens, and for Sonnet 5.5 it’s $2 and $10.

Watch for: you pay per token here, whatever Claude plan you have. OpenRouter’s Claude Code guide says “Claude Code with OpenRouter is only guaranteed to work with the Anthropic first-party provider,” and its post adds that “Latency and uptime can still differ between providers.” The by-hand setup also says “If you were signed in with a Claude account, run /logout once,” the one thing Anthropic asked people not to do during September’s sign-in incident. Set it up before you need it.

Codex, on a ChatGPT plan

Best for: a second lab’s agent, on a plan you may already have.

Codex is OpenAI’s coding agent, and it runs on OpenAI’s service, not Anthropic’s. Check status.openai.com before you move. Codex’s pricing puts it in every ChatGPT plan, from Free.

It can pick up Claude Code’s work from your own disk. OpenAI’s import page says to “Type /import in an idle Codex CLI session,” choose Claude Code, and “Codex CLI imports up to 50 chats from the last 30 days.” The ChatGPT desktop app has the same under Settings > Import, and it turns Claude’s instruction files into AGENTS.md. “Importing doesn’t change or delete your existing agent setup.” From inside Claude Code, OpenAI’s Codex plugin has /codex:transfer, which “Creates a persistent Codex thread from the current Claude Code session and prints a codex resume <session-id> command.”

Price: Free is $0, Go $8 a month, Plus $20, and Pro “Plans at $100, $200, or $500 USD per month.”

Watch for: Free and Go get “GPT-6 Luna at Standard speed in the desktop app, subject to rollout,” not OpenAI’s top models. OpenAI says to “Review imported setup before you rely on it,” especially “Hooks whose behavior may differ after import.” Install Codex and sign in before the day you need it. Our guide to using the two together covers the rest.

A local model, through Ollama

Best for: work that can’t depend on anyone’s servers, on a computer with memory to spare.

“Ollama connects Claude Code to local and cloud models through its Anthropic-compatible API,” says its guide, and one command starts it:

ollama launch claude

You stay in Claude Code, and the model runs on your own computer. Ollama lists file edits, tool calling, subagents and running commands “with Claude Code’s permission flow” among what works.

Price: free. Ollama’s pricing starts at $0, which includes “Run models locally.”

Watch for: Ollama says to “Choose a model with enough context for your repository,” and for local ones to “set a 64k+ context window.” Tool calling works “with compatible models,” so not every model can drive Claude Code. Download the model before an outage, not during one. Our guide to local models covers which to try and what they need.

T3 Code

Best for: switching agents inside one thread, free and open source.

T3 Code calls itself “The open-source control plane for coding agents”: “Orchestrate Claude Code, Codex, Antigravity, OpenCode, Cursor, and Grok from one surface. Bring your own subscription.” When one provider is down, you switch the thread to another, and T3 Code “transfers conversation context when you switch providers.” Its FAQ says it works on Windows, Mac and Linux.

Price: free. The FAQ says “There’s no way to pay us to use T3 Code. You bring your own sub. It is free and open source (MIT).”

Watch for: “A handoff does not copy the outgoing provider’s reasoning, tool-call state, or attachments,” and for a long conversation it’s “a budgeted selection, not an agent-written summary.” Its README says “We are very very early in this project. Expect bugs.” You still need a second agent installed and signed in.

Enjoy

Best for: handing the conversation you’re in to another lab’s agent, and back to Claude when it returns.

Enjoy runs Claude Code, Codex, Grok Build, Antigravity, Muse Code, and local models through OpenCode, on Mac, Windows and Linux, on the plans you already have. When Claude Code can’t finish a request:

• You see why. The conversation’s status changes to Failed, and Enjoy shows “a one-line reason, the agent’s own error under Details, and Retry.”
• Another agent retries it. Change the agent in the message box, and “a second button appears, starting with Retry with and naming what changed.” The new agent “gets the conversation so far, including replies from other agents.” Changing only the model works the same way, for an incident on one model.
• Or a local model does. Switch to OpenCode with a model on your own computer.
• Claude picks it up later. “Each agent keeps its own place in a conversation. When you switch back to an agent, it continues where it left off.”

Price: Enjoy is one monthly subscription, and the first seven days are free, with a card to start. It includes your phone and any browser and one computer; each other computer is $9.99 a month, and each teammate you invite needs a subscription of their own. See pricing.

Watch for: Enjoy doesn’t bring Anthropic’s service back, and it doesn’t switch agents for you when one fails: the request waits until you choose. The second agent has to be installed and signed in, on a plan of its own or a free one, and that’s quicker done before an outage. Enjoy isn’t open source, and it’s a monthly subscription on top of your AI plans once the free week ends.

When to stay with Claude and wait

Most of the time. If the status page shows an incident on one model, switch models and carry on. If it’s wider, the last 30 days say it will most likely be over within the hour, and Claude Code retries in the meantime. Setting up a cloud account or a gateway in the middle of an outage can take longer than the outage does.

Set up a second way when an hour matters: before a deadline, for a script that runs unattended, or when you already pay for ChatGPT. The cheapest insurance is free: Codex signed in on the ChatGPT plan you have, or a local model already downloaded.

A bigger Claude plan isn’t a documented fix. Claude’s pricing lists “Priority access at high traffic times” on Max, and Ultra includes everything in Max. But a 529 means the API is “at capacity across all users,” and neither the pricing page nor Claude Code’s error reference says how the two fit together.

And don’t treat a limit as an outage. “You’ve hit your session limit” comes with a reset time, and our usage-limit guide compares the ways through it.

How we compared

We read Anthropic’s status page and its incident history, Claude Code’s docs on errors, model configuration, its command line and its cloud providers, Claude’s pricing, OpenRouter’s post, guide and pricing, OpenAI’s Codex docs, plugin and status page, Ollama’s guide and pricing, T3 Code’s site and docs, and Enjoy’s own help on November 11, 2026, and every one is linked below.

The incident counts are ours. We counted the incidents on Anthropic’s status page that list Claude Code as affected and were posted from October 12 to November 10, 2026, by UTC time, and measured each from its first post to the post that resolved it. A four-minute incident on November 10, “Overloaded errors on Fable 5.2 and Mythos 5.2,” lists no product, so it isn’t in the count.

We quote each maker’s own words for its limits and leave out what it hasn’t documented. Status pages and plans change often, so we recheck this guide against its sources at least once a month, and the date at the foot of the page says when we last did.