wtf is antigravity-sdk-python?

The Google Antigravity SDK is a Python toolkit for building AI agents powered by Google's Gemini model. It handles the heavy lifting of agent infrastructure, managing state, tool calls, and conversation history, so you can focus on what your agent actually does rather than the plumbing required to run it. At its core, you create an agent, give it instructions, and chat with it. The agent can call tools you define (like a Python function that checks weather), ingest multimedia files like PDFs and images alongside text, and stream responses back in real time. It also supports connecting to external MCP (Model Context Protocol) servers for additional tool access. A built-in policy system lets you control what the agent is allowed to do, for example, blocking all file writes by default but allowing reads, or prompting a human for confirmation before running commands. Triggers let you set up background tasks that periodically feed messages into the agent, useful for things like monitoring deployment status on a schedule. This is aimed at developers building AI-powered assistants, coding tools, or automated workflows who want a structured framework rather than wiring raw model calls together. A practical example from the docs is an agent that navigates a codebase, answers questions about files in the current directory, and can read specifications or analyze architecture diagrams, all within a single conversation that remembers its history. One thing worth noting is that this SDK depends on a compiled runtime binary that ships with the published package. Cloning the repository alone won't work, you need to install it via pip to get that binary. The project is organized in three layers: a simplified entry point for quick starts, a session layer for managing ongoing conversations, and an adapter layer that abstracts the underlying transport. This tiered design means you can start simple and dig into lower layers as your needs get more complex.