I’m working on a project where I’ll be tracking finances and inventory via Antigravity. Everything was going very well until this week, but now I keep getting this error. There are no MCP servers or any skills involved. I’ve connected to and then deleted MCP servers before. Could you guide me on how to check and delete them? But how do I get rid of this error? Can you help me?
I have the same problem. What’s going on?
I pay for the Pro Plan.
Unfortunately, there’s no change; this error is very strange.
Hi @Adem_OZTURK,
The Agent terminated due to error is a generic failure message that can occur for several reasons, including temporary model capacity issues (HTTP 503) or corrupted local session data. You can try the following workaround and let us know if it resolves
-
Even if you are not using MCP server, stale entries in your configuration file can sometimes cause the agent to fail during its initialization phase. You can manually delete them in
mcp_config.jsonfile. Open this file with a text editor and ensure themcpServersobject is empty, like this:{"mcpServers": {}}. Save the file and restart Antigravity -
Switch to a different model - Use the model selector in the chat pane to switch to a Gemini model (e.g., Gemini 3.1 Pro) or vice versa and try your prompt again
-
If the error is persistent across all conversations, your local application cache may be corrupted. You can delete follwing folder to reset local state
- Windows:
%APPDATA%\Antigravity - macOS:
~/Library/Application Support/Antigravity
- Windows:
-
If you are working with large inventory, the current conversation history might be hitting a context limit. Starting a
New Conversationoften resolves this immediately
Hi @Titan,
Welcome to the Forum,
Please note that the error is a technical failure message and is usually not related to your plan.
It most often occurs when a specific model is experiencing temporary capacity issues or when local session data becomes corrupted.
Since you are on the Pro Plan, I recommend these steps to resolve the issue:
-
Even with Pro access, certain models may reach capacity during peak times. Use the model selector to switch to a Gemini model (e.g., Gemini 3.1 Pro) and see if the error persists
-
Sometimes the
Prostatus needs to be re-synced with the local client. Log out of Antigravity and Log back in to refresh your authentication token -
If the error happens on every prompt across all models, your local cache is likely corrupted. You can delete follwing folder to reset local state
- Windows:
%APPDATA%\Antigravity - macOS:
~/Library/Application Support/Antigravity
- Windows:
-
If a specific thread has become too long or contains very large files, it can trigger a termination error. Starting a New Conversation often clears the stuck state
- this did not work, i restored previous conversation summary into new conversation window.
- requested to turn a markdown into a artifact, i am getting agent terminated due to error
i also deleted ~/Library/Application Support/Antigravity which restarted everything, including onboarding process.
i am still getting agent terminated due to error.
I managed to resolve this issue.
In my case, the problem wasn’t related to MCP servers or skills directly, but rather an internal configuration/state issue within the environment. I worked through it with another LLM, which helped identify and clean up the conflicting setup.
That said, I’m no longer using Google Antigravity, so I haven’t investigated it further within this ecosystem.
If you’re facing the same error, I’d recommend checking for any leftover or partially removed configurations and making sure the environment is fully clean before reinitializing.
Hope this helps someone ![]()
This could happen if too many skills are initialized, such as Antigravity awesome skills 1300+

