google-antigravity/antigravity-sdk-python

Builds AI agents in Python against Google's Antigravity infrastructure, which means you get access to the full platform surface rather than a thin wrapper around a single model endpoint. That distinction matters: most agent SDKs abstract away so much that you hit a ceiling the moment your workflow needs anything non-standard. This one appears to expose the low-level primitives, which gives a technical founder actual control over how agents are composed, how state is managed, and what hooks exist for observability. The repo has a high signal score for a newly published open-source project, which suggests real developer interest rather than a launch-day spike. Reservation: Google has a track record of deprecating developer-facing products without warning, so betting a production workflow on this before it reaches stable API status carries genuine risk. Worth cloning and reading the architecture before committing. -> Best for: AI engineer building multi-step agent workflows on Google infrastructure