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

Disclosure: Enjoy is our product, and it runs Antigravity. We checked every way below against its maker’s own pages on November 12, 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 Antigravity stops answering, the first problem is telling whether it’s down at all. Google doesn’t publish a status page for it. Antigravity’s support page links to its docs, X, YouTube and LinkedIn, and Google’s status pages for AI Studio and Google Cloud don’t list it. The fullest record is Google’s own AI developer forum, where a thread on the high-traffic error, opened on April 8, 2026, has more than 400 posts and about 61,000 views.

Google’s plans page says why capacity runs short: “The baseline rate limits are primarily determined by available capacity,” and “These adjustments may be necessary to manage system capacity and maintain service stability.” Many of the forum’s reports name one model failing while others work, so the first two ways below cost nothing and are often enough. The rest are for the long outages, and for work that can’t wait.

Way	Cost, checked November 12, 2026	How soon you’re back	Best for
Retry, and let Antigravity retry	Free	When Google’s capacity returns	A short outage
Switch models	Included in your plan	Right away, if the other model has capacity	An outage on one model
Antigravity CLI with a Gemini API key	Gemini 4 Hydrogen free on the API’s free tier; paid, $1.50 in and $7.50 out per million tokens	Right away, once it’s set up	Trouble with your account, not Gemini
Antigravity on a Google Cloud project	Gemini Enterprise from $30 a seat a month, or billed by use	Once an administrator sets it up	Teams that already have Gemini Enterprise
Claude Code, on a Claude plan	Pro $20 a month; Max $100; Ultra $200	Right away, in a new conversation	Another lab’s agent, on its own servers
Codex, on a ChatGPT plan	Free; Go $8 a month; Plus $20; Pro from $100	Right away, in a new conversation	A second agent at no cost
OpenCode, with free or local models	Free	Right away; a local model 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, with a free first week (pricing)	Right away, in the same conversation	Handing the conversation to another lab’s agent

Prices are each company’s US list prices, before tax. Retrying, switching models, the Gemini API’s free tier, Codex on ChatGPT’s Free plan, OpenCode’s free models and a local model cost nothing extra.

First, read the message

Not every stop is an outage, and Google’s docs don’t list Antigravity’s errors in one place. These are the ones that come up in Google’s changelog and in the reports on its forum:

What Antigravity shows	What it means	What to do
Our servers are experiencing high traffic right now, please try again in a minute.	A model is out of capacity on Google’s side. In the CLI, people report it with UNAVAILABLE (code 503): No capacity available for model and the model’s name	Try again in a minute, or switch models
Model unavailable (attempt 2/10), retrying in 12s	Antigravity CLI is “retrying an unavailable model,” with a countdown	Let it count down, or switch models
Agent execution terminated due to error.	The general error for a turn that failed; the cause varies	Select Retry, then try another model
HTTP 429 RESOURCE_EXHAUSTED	A rate limit or your quota	Check View Usage in the model picker
Your AI credits balance is too low to continue.	Your quota and your AI credits ran out	See our quota guide
Working… with no reply	In one forum report, the app sat there while the CLI showed the 503 for the same model	Stop it, switch models, and send again

“Agent execution terminated due to error.” covers a lot. Before fixes in Antigravity CLI, Google’s changelog shows it ending turns after “laptop sleep or a Wi-Fi change,” and after AI credits ran out, “instead of retrying on Working... for about two and a half minutes.” So check your network and your quota before you blame Google. A 429 can be either kind: the CLI retries “per-minute 429 rate limits” by itself, and stops “right away” when “the quota is a daily or billing cap.” When the model picker’s View Usage shows quota left and every request still fails, it’s Google’s capacity, not your plan.

Retry, and let Antigravity retry

Best for: a short outage.

Antigravity retries before it shows you an error. The Antigravity App’s changelog for version 2.14.0, on September 15, 2026, says “Transient errors are now retried with backoff for about twelve minutes.” When it gives up, the error card has Retry, which since version 2.9.1 “resumes the task directly, without adding an artificial ‘Continue’ message to your conversation.”

Antigravity CLI retries 502, 503, 504 and per-minute 429 errors “with exponential backoff,” and now “waits at most 30 seconds between automatic retries instead of up to 4 minutes.” Its status line shows what it’s waiting on, as in Model unavailable (attempt 2/10), retrying in 12s. If you close it, agy -c or agy --continue resumes “threads instantly based on your active workspace,” per its conversations docs.

Price: free.

Watch for: Google posts no incidents, so nothing tells you how long one will last. Some last days on one model. In a November 5 report, Gemini 3.8 Flash had been “completely unusable” for “the past 2+ days,” while “Gemini 4 Argon Low works fast,” as the post’s title put it. Retries only wait out the short ones.

Switch models

Best for: an outage on one model, which many of the forum’s reports describe.

Antigravity’s models page says “You can switch models at any time,” from the model selector under the prompt box, or with /model in the CLI. The reports show why it helps. On August 14, one user’s Gemini and GPT models failed while “Claude works normally.” On November 11, another’s Gemini 4 Argon (High) failed on every prompt in the app, the CLI and the IDE, with quota left. The 503 names the model with its thinking level, as in gemini-3.8-flash-high, so another level of the same model may answer when that one doesn’t. Google hasn’t documented how it divides capacity.

What you can switch to depends on your plan. On Antigravity’s models page, Gemini 4 Hydrogen and Gemini 3.8 Flash are on every plan, though Google’s announcement brings Hydrogen to paid Pro and Ultra, as our memo explains. Free, Google AI Plus and Pro also have Claude Sonnet 4.6 and Opus 4.6, and every plan but Enterprise has GPT-OSS-120b, until each “Will be removed on November 18, 2026.” Paid Pro adds Gemini 4 Argon and the Claude 5.5 models, and Ultra adds Gemini 4 Carbon and Orchestrator.

Price: included in your plan.

Watch for: a switch waits for the turn in progress: “If you change the reasoning model while the agent is running, it continues to use the previously selected reasoning model until it completes its steps for that turn (or until you cancel the current execution).” Quota comes in two groups, “Gemini Models” and “Claude and GPT models,” so a switch can run into a limit instead. And some outages take every model: the day the high-traffic thread opened, one of the first replies said “All models (including Claude, Sonnet, Pro, and non-Pro) are currently not responding.”

Antigravity CLI with a Gemini API key

Best for: keeping Antigravity’s agent when the trouble is your account or Antigravity’s own service, not Gemini.

Google’s authentication docs say you can “Run Antigravity CLI with your own Gemini API key instead of a signed-in Google account. Model requests go directly to the Gemini API, and the CLI never establishes an account session.” Create a key in Google AI Studio, set "modelProvider": "gemini" in ~/.gemini/antigravity-cli/settings.json, then:

export GEMINI_API_KEY="your-api-key"
agy

Both steps matter: “Only setting a GEMINI_API_KEY environment variable on its own has no effect.” Unlike Antigravity itself, the API has a status page, at aistudio.google.com/status.

Price: on the Gemini API’s pricing page, Gemini 4 Hydrogen and Gemini 3.8 Flash are “Free of charge” on the free tier. On the paid tier, Hydrogen is $1.50 per million input tokens and $7.50 per million output, and 3.8 Flash is $0.75 and $3.75 through December 31, 2026.

Watch for: it’s Gemini only, and in the CLI only: Google documents the key for Antigravity CLI and says “This suits headless and CI runs.” On the free tier, your content is “used to improve our products,” and the paid tier’s isn’t. And the API has bad days too. On October 30, AI Studio’s status page posted a “Major outage affecting API key traffic in AI Studio”: “All requests using API keys are currently failing.” It was detected at 11:42 and resolved at 14:18.

Antigravity on a Google Cloud project

Best for: teams that already have Gemini Enterprise.

Antigravity can sign in through Google Cloud instead of a personal Google account. Its enterprise docs give two ways in: “a Google Cloud project with an invoiced Cloud Billing account” for “consumption-based billing,” or “your Gemini Enterprise Standard or Gemini Enterprise Plus seat license.” Turning on the Gemini Enterprise API is what allows “Antigravity clients to connect to your project’s model endpoints,” and Google Cloud’s status page lists Gemini Enterprise, so on this route there’s a status page to check. For scripts, Antigravity CLI can use Application Default Credentials:

gcloud auth application-default login --project {GCP_PROJECT}
export AGY_ADC_AUTH=true

Price: Gemini Enterprise Standard and Plus start at $30 a seat a month, per Google’s plans, or your project pays for what it uses.

Watch for: this is set-up work for an administrator, not a fix for this afternoon. It needs an IAM role, roles/discoveryengine.agentspaceUser, and the Gemini Enterprise API turned on in the project. “Antigravity IDE (standalone) is not supported for enterprise deployments,” and with ADC, “models older than Gemini 3 Flash are not supported.” Google doesn’t say this route stays up when personal accounts don’t.

Claude Code, on a Claude plan

Best for: another lab’s agent, on Anthropic’s own servers.

Claude Code is Anthropic’s coding agent, and an Antigravity outage doesn’t reach it. Check status.claude.com before you move. Claude’s pricing says “Claude Code is included in all paid plans.”

Price: Pro at $20 a month, or $17 a month billed yearly ($200 up front), Max at $100 and Ultra at $200. The free plan leaves Claude Code out.

Watch for: it starts a new conversation and doesn’t see Antigravity’s, so tell it where things stand. Anthropic’s limits replace Google’s: “Your activity across Claude on web, desktop, mobile, and Claude Code all draws from the same pool.” And don’t sign it in with your Google account: Antigravity’s FAQ says “Using third-party software, tools, or services to access Antigravity is a violation of our Terms of Service.” Our Claude Code outage guide covers Anthropic’s bad days.

Codex, on a ChatGPT plan

Best for: a second agent at no cost, on a plan you may already have.

“ChatGPT Work and Codex are included in your ChatGPT Free, Go, Plus, Pro, Business, Edu, or Enterprise plan,” says Codex’s pricing. It runs on OpenAI’s service, so check status.openai.com.

Price: Free is $0, Go $8 a month and Plus $20, and Pro comes in “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. Codex’s import brings in work from Claude Code, Claude Cowork or Cursor, not Antigravity, so it starts fresh too. Install it and sign in before the day you need it. Our Codex outage guide covers OpenAI’s.

OpenCode, with free or local models

Best for: work that can’t depend on anyone’s servers.

OpenCode is an open-source coding agent for the terminal, the desktop and more. Its homepage says “Start with free models or bring any of 200+ providers,” and “OpenCode comes with a set of free models. No account needed.” To leave every server out of it, Ollama runs a model on your own computer, and its guide starts OpenCode with one command:

ollama launch opencode

Price: free. Ollama’s pricing has a $0 plan that includes “Run models locally.”

Watch for: OpenCode’s free models are free “for a limited time,” and for some, “The team is using this time to collect feedback and improve the model.” A local model needs memory to spare: “OpenCode requires a context length of 64k or higher,” says Ollama, and it’s a download to do before an outage, not during one. Don’t sign OpenCode in with your Antigravity login: Google’s FAQ names it, and says that “can result in suspension or termination of your account.” Our guide to local models covers which to try.

T3 Code

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

T3 Code says it can “Orchestrate Claude Code, Codex, Antigravity, OpenCode, Cursor, and Grok from one surface. Bring your own subscription.” Antigravity is one of its agents, so when it fails you switch the thread to another, and T3 Code “transfers conversation context when you switch providers.” It runs on Mac, Windows and Linux.

Price: free. Its FAQ says “There’s no way to pay us to use T3 Code. You bring your own sub.”

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.” T3 Code reaches Antigravity through Google’s service like every other app, so it can’t bring Antigravity back, and 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 Antigravity when it recovers.

Enjoy runs Antigravity on your Google account, through the Antigravity server Google offers to apps such as Zed and JetBrains IDEs, along with Claude Code, Codex, Grok Build, Muse Code, and local models through OpenCode, on Mac, Windows and Linux. When Antigravity 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 outage on one model.
• Or a local model does. Switch to OpenCode with a model on your own computer.
• Antigravity 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 week is free, with a card to start. It includes your phone and any browser and one computer; each other computer costs extra each month, and each teammate you invite needs a subscription of their own. See pricing.

Watch for: Enjoy doesn’t bring Google’s service back: Antigravity in Enjoy uses the same service and fails with it. Enjoy doesn’t switch agents for you when one fails, either: the request waits until you choose. It uses one Google account for Antigravity in every project. 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 on Antigravity and wait

When one model fails, switch models and carry on. When they all do, Antigravity is already retrying, and the app keeps at it for about twelve minutes. Google posts no incident history, so there’s no telling how long an outage will last, and setting up an API key or a Cloud project in the middle of one 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 Claude or ChatGPT. The cheapest insurance is free: a Gemini API key saved for Antigravity CLI, Codex signed in on ChatGPT’s Free plan, or a local model already downloaded.

A bigger Google plan isn’t a documented fix. Google’s plans give Ultra “Higher rate limits” and “Highest rate limits” in Antigravity, but high-traffic errors come with quota left, and Google doesn’t say whether paying more puts you first in line for capacity.

Skip the fixes that sign other tools in with your Google login or rotate accounts for you. Google’s FAQ says “Wrapping agy inside third-party coding agents or applications, or automating account rotation, is not supported under consumer plans.” And don’t treat a quota as an outage: our quota guide compares the ways through that.

How we compared

We read Antigravity’s support page, its docs on models, plans, authentication, Gemini Enterprise, CLI conversations and its FAQ, and its changelog; Google’s AI Studio and Google Cloud status pages, the Gemini API’s pricing and Google’s AI plans; the threads on Google’s AI developer forum linked above; Anthropic’s and OpenAI’s pricing and status pages and OpenAI’s import docs; OpenCode’s homepage and free models; Ollama’s guide and pricing; T3 Code’s site and docs; and Enjoy’s own help, on November 12, 2026. Every one is linked below.

Google publishes no incident history for Antigravity, so unlike our guides to Claude Code and Codex outages, this one has no count of incidents or how long they lasted. The forum’s reports are users’ own, and we say so where we quote them; the post and view counts are the forum’s on November 12. We quote each maker’s own words for its limits and leave out what it hasn’t documented. 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.