Today I was confused by Antigravity errors about high load on their services. It made my work impossible even with the cheapest model Gemini 3 Flash.
Some time ago I heard something about the OpenCode. And it was the time to try it!
I've installed the opencode in my system by brew install anomalyco/tap/opencode and respective extension from the marketplace.
I have a good documentation inside the repo like described in the article Specification-First Agentic Development: A Methodology for Structured, Traceable AI-Assisted Development. So the default free OpenCode model Big Pickle performed planning, reviewing and coding stage well.
But then I realized that it was working without taking into the account system instruction and rules which I had for Antigravity.
So I've performed calls of Antigravity assurance workflows (like here and here) right from the OpenCode chat and it performed them perfectly.
As I have a lot of workflows for linting, security check of diff and the whole repo, and especially external self-made security gateway I was sure that the quality of code produced by the OpenCode was good enough and aligned with my codebase.
The only thing I can mention is a redundant file was left after some iterations of testing. But it can be fixed by a good review right after MR creation.
So seems the OpenCode is a good fallback for cases when Google servers are experiencing problems. Also it can by used to save tokens for some kind of tasks.
Top comments (3)
Smart approach—using OpenCode as a fallback not only keeps workflows resilient during outages but also highlights the real power of multi-agent flexibility in modern AI-driven development 🚀
Yeah! I'm already thinking about running them simultaneously for different tasks or to check each other 😁
Nice—that kind of parallel + cross-checking setup could really boost reliability and catch edge cases early, solid thinking 😄