Hi all — quick heads-up based on recent investigation.

It seems quota/rate-limit enforcement has become noticeably stricter lately, so it’s expected that we may hit rate limits more often than before (even when quota still looks “available”). In my pool, some accounts also end up effectively unusable for long periods once they start rate limiting, while a few others continue working normally.

Also worth noting: OpenCode can end up calling the endpoint more aggressively than a native/chat UI (tool calls, retries, streaming, multi-turn chains), which makes it much easier to trigger rate limits compared to “normal” usage patterns.

Practical guidance:

If you only have one account, you’ll likely have a better experience using Antigravity IDE directly instead of routing through OpenCode, since OpenCode’s request pattern can trigger limits faster.
If you do use OpenCode, expect rate limits and consider reducing request volume / burstiness (shorter sessions, fewer parallel/retry-heavy workflows, etc.).
If others are seeing similar behavior, please share status codes and how long the rate-limit state persists so we can better characterize it.

Another clarification:
Using this plugin may increase the chance of triggering automated abuse/rate-limit protections due to higher request volume (retries, tools, streaming, multi-turn chains). The upstream provider can restrict, suspend, or terminate access to an account at their discretion if activity is flagged or violates their terms—so use at your own risk, keep request volume reasonable, and don’t rely on a single account for critical usage.
USE IT AS YOUR OUR RISKS