The Three Pillars of Antigravity SDK
The Google Antigravity SDK orchestrates autonomous AI workflows using a hierarchical architecture:
Agent
The entry point for configurations (model selection, capabilities, tools list, safety policy predicates). Abstracting away environmental transport details.
Conversation
Manages the session state, accumulation of step histories, context compaction rules, and supports messaging loops (`agent.chat()`).
Connection
Handles underlying transports (Stdio, SSE, or Fetch API gateways) to forward payloads to local or remote serverless instances.
Domain Playbook Skills (Agent Knowledge Files)
Skills allow capturing specialized expertise and repeatable workflows inside filesystem directories containing standard `SKILL.md` documents.
google-antigravity-sdkInstantiates agent config paths and sets up model parameters.
a11y-debuggingPerforms accessibility audits on DOM element positions and contrast ratios.
pubmed-databaseQueries scientific biomedical literature to find experimental data and citations.
clinical-trials-databaseRetrieves recruiting trial details and NCT sponsor portfolios.
pymolRenders 3D macromolecule protein structures and alignments.
chembl-databaseQueries bioactive molecules, approved drugs, and IC50/Ki constants.
Netlify Functions Fetch REST API Specs
AI agents connect to physical platforms using Netlify serverless fetch APIs. Below is the REST schema specification:
1. Get Schema Specification
2. Post Force Command