Antigravity on macOS starts the language server but fails internal bootstrap

Google Antigravity
1

/proxy/unleash/frontend returns 404 and GetAgentScripts returns 500, leaving chat/agent non-functional

Severity: High / workflow blocking

Environment:

  • OS: macOS

  • App: Antigravity desktop app

  • Account: Pro/paid account

  • Workspace impact: Happens even after local cleanup and extension isolation tests

Description:
I am experiencing a blocking startup issue in Antigravity on macOS. The app launches and the language server appears to start successfully, but the internal initialization sequence does not complete. After startup, prompts either receive no response or execution produces no usable output.

The failure pattern is consistent and appears to be internal to Antigravity rather than related to a specific project or prompt. In the developer console and extension host logs, I consistently see errors indicating that the Unleash provider is not initialized, local feature-toggle requests return 404, and a language-server endpoint for agent scripts returns 500. Similar public reports describe this same pattern as a startup/bootstrap failure that can leave chat unavailable, break session initialization, and prevent agent functionality from loading correctly.

Observed behavior:

  • Antigravity starts and logs that the language server started successfully.

  • Immediately after, internal initialization errors appear.

  • The chat/agent becomes unreliable or unusable.

  • Prompts may hang silently or return no useful execution result.

  • Local internal requests fail during startup.

Relevant log lines:

[Extension Host] Error retrieving Unleash sample rate Error: UnleashProvider must be initialized first!
(Antigravity) [INFO]: Language server started
(Antigravity) [INFO]: LS lspClient started successfully
POST https://127.0.0.1:<port>/proxy/unleash/frontend 404 (Not Found)
Unleash: Fetching feature toggles did not have an ok response
[AntigravityUnleashService] [object Object]
POST https://127.0.0.1:<port>/exa.language_server_pb.LanguageServerService/GetAgentScripts 500 (Internal Server Error)

Steps to reproduce

  1. Launch Antigravity on macOS.

  2. Wait for the app to finish startup.

  3. Open Developer Tools / Extension Host logs.

  4. Observe that the language server reports successful startup.

  5. Observe that internal bootstrap then fails with UnleashProvider must be initialized first!, repeated POST /proxy/unleash/frontend 404 responses, and GetAgentScripts 500.

  6. Attempt to use chat or run a prompt.

  7. Observe that the agent becomes silent, unstable, or fails to return expected execution results.

Public reports describe closely related startup failures where Antigravity tries to use internal services before initialization is complete, especially around LanguageServerClient or UnleashProvider, and those failures can leave chat or agent features unavailable.

Other reports specifically show that POST https://127.0.0.1:<port>/proxy/unleash/frontend returning 404 is associated with broken session/bootstrap behavior and missing feature-toggle.

Expected behavior

Antigravity should complete its startup sequence cleanly, initialize internal providers before using them, successfully fetch required feature-toggle/configuration data, and allow chat/agent execution to work normally after launch.
Related reports indicate that when initialization succeeds, the extension host should not attempt to access singleton-style services before they are ready, and the local internal endpoints should respond normally rather than failing with 404 or 500.

Actual behavior

The language server appears to start, but Antigravity immediately hits internal initialization failures tied to Unleash/bootstrap and then fails downstream when calling GetAgentScripts.
This leaves the app partially started but functionally broken for chat/agent use.
Recent public reports describe the same general symptom cluster as internal bootstrap failure, broken feature flags, missing services, unavailable chat, and agent execution failures across models.

Troubleshooting already performed

I already performed local cleanup steps and the issue persisted.
I cleared local application state/cache and retried startup.
I also tested with reduced extension interference / a cleaner startup path, but the same Antigravity-specific errors remained.

This matches other reported cases where wiping local app data, clearing extension/cache state, or retrying with a fresh environment did not resolve the underlying Unleash/bootstrap failure.
There are also public reports where multiple features stayed broken after a clean reinstall because the initialization chain still failed upstream at Unleash/feature-toggle startup.

Suspected root cause

This looks like an internal initialization-order or bootstrap bug inside Antigravity.
Based on similar reports, a likely failure chain is: Unleash/feature-toggle initialization fails, the provider is accessed before it is ready, startup continues in a partially initialized state, and later language-server calls such as GetAgentScriptsfail as a consequence.
The repeated 404 on /proxy/unleash/frontend and the 500 on GetAgentScripts strongly suggest this is not a prompt-content issue, but a product-side startup or routing failure in Antigravity’s local backend/bootstrap flow.

Impact

This issue is fully blocking normal use of Antigravity for chat and agent execution.
It prevents reliable prompting, breaks execution feedback, and makes the desktop app unusable for development workflows.