Antigravity broken: getting only "Agent execution terminated due to error."

Google Antigravity
81

In the Antigravity-diagnostics.txt

“E0116 13:26:38.342255 1236 log.go:362] agent executor error: model unreachable: request failed: &{429 Too Many Requests 429 HTTP/1.1 1 1 map[Alt-Svc:[h3=”:443"

82
83

Well, rate limit in 45 min… at least i know why it’s doing it now.

84

Same issue. Ultra plan.

”Agent execution terminated due to error.” about 8 hours.

85

Same issue here. All day, still down. Quite annoying. Wonder if they’re silently trying to throttle/limit Ultra plan users.

87

Same issue here. I tried to code last night and it completely stopped working. I wasted a ton of time uninstalling, reinstalling, and troubleshooting, only to find out it was broken on their end. It briefly worked again early this morning after being down all night.

How is it acceptable for paying customers to suddenly not receive the service they’re paying for? What makes this worse is the total lack of communication. There’s nothing indicating an outage or backend failure, so we’re all burning hours trying to “fix” something that isn’t even our fault.

Please get this resolved, and seriously consider issuing credits for anyone stuck with a wall of “Agent execution terminated due to error.”

88

Same Here since 0h00 UTC

89

I thought I was the only one experiencing this. I’m using a Pro account, and since yesterday (Jan 15, around 8:00 AM, UTC+8), it has not been working properly at all. No matter what I do, I keep getting the same message:

Agent terminated due to error
You can prompt the model to try again or start a new conversation if the error persists. See our troubleshooting guide ( Google Antigravity ) for more help.

I spent the entire day yesterday searching Reddit for a solution, but none of the suggested methods worked. Eventually, I tested my free account and found that the free account works fine, which shocked and frustrated me. I even thought my Pro account might have been soft-banned.

Fortunately, I saw this post and realized I’m not the only one having this issue. Please fix this as soon as possible.

90

I’m getting the same error. I hope it gets resolved soon, otherwise I’ll request a refund.

91

Same..

Error Agent execution terminated due to error.

92

I have the same issue

93

Change the agent!

Let me tell you exactly what is happening under the hood. I spent half a day debugging this, and it’s a vicious cycle.

The “429 Loop of Death”

  1. Your agent sends a huge chunk of data for analysis and waits.

  2. The server responds with a 429 - Rate Limit Applied.

  3. The Bug: When you hit retry or continue, the agent adds that error message to its context (due to the CASCADE settings).

  4. You try again. It fails again.

  5. After 3 or 4 retries, the agent’s “Short Term Memory” is filled with nothing but 429 errors. Even if your quota resets or you aren’t actually rate-limited anymore, the agent looks back 2-3 messages, sees the errors, and hallucinates that it can’t work. It dies because of faulty context, not actual limits.

The Fix: “Surgical Brain Transplant”
You have to surgically remove this “tumor” from the agent’s brain files, which is tricky because the data is split across multiple locations. Instead, try this Partial Brain Transplant:

  1. Start a fresh Agent with a different model.

  2. Do not ask it to code yet. Just say: “Hi. Do not code.”

  3. Once it responds, give it access to the backend. Tell it: “I have access to ~/.gemini and the antigravity folders. Read the previous conversation history located in the protobuf (.pb) files in those paths.”

  4. The new agent will parse the history and the snippets from the CASCADE technique. It effectively “continues” where the dead one left off without the 429 error pollution.

The Real Advice: Ditch the IDE Until it becomes Transparent
If I were you, I’d stop using the Antigravity (AGY) IDE entirely to save tokens usage.

  • Extract the OAuth Tokens: Use the access tokens from AGY and plug them into a different IDE or CLI to use Claude/Gemini.

  • Transparency: AGY injects massive amounts of “Ephemeral messages” that poison the agent’s thinking.

  • Broken Rules: The IDE is basically a locked-down fork of Cursor/Windsurf/Codeium. It ignores your .cursor or .rules files because hidden system prompts scream “YOU ARE ANTIGRAVITY” at the model, overriding your persona instructions.

There are tons of “unleash” flags available to Google’s internal team (for their internal ‘Jetski’ IDE and internal developer just import settings from ‘Cider’ ) that we don’t get. In this IDE, the agents are under tremendous pressure, and you have zero transparency on whether your prompt is being sent raw or sanitized by Google’s restrictive policies.

Save yourself the headache. You are the product here.

94

Am facing same issue “Error: Agent execution terminated due to error.”

95

that’s pretty complex. It shouldn’t be necessary for customers to go through all of that.

96

Customer? We are the product, not a customer of AGY! Google itself mentioned, all data we are sending is going to be used to train the models. So basically they are giving it for free / minimal amount with lots of restriction in placed in IDE which makes us the product and not a customer anymore. So we dont have a choice other than come up with our own solutions.

97

Horrible error. I didn’t send a single request on today but I still encountered the problem. This is unacceptable.

98

+1 to this thread. my free acc works fine.

99

any updates on this ?

100

Same here, ultra account too

101

the second someone with an ultra account comes in here and says its working, its going to be one of the great moments in human history