Perform 3-axis precision detumble fusing RCS thrusters and reaction wheels to eliminate tumbling across pitch, yaw, and roll axes.
Let’s build a smart flight computer that can survive in deep space. Instead of writing fixed scripts to solve missions, modern flight computers operate in an autonomous loop: Observe, Orient, Reason, and Act. Rather than following hardcoded control rules, an autonomous agent continuously observes telemetry, orients itself within the flight envelope, reasons through trade-offs, and acts to achieve mission objectives.
In this workshop, you’ll use the SDK. The Antigravity SDK lets you build custom agents by connecting code directly to Google's agent harness. It’s the same harness powering Antigravity 2.0, the Antigravity IDE, and many of Google’s internal agents.
Open your project directory and prompt Antigravity:
Read the sample code in this directory. We’ll be completely replacing it.
Build an agent that can play missions T0-01, T0-02, and T1-01.
Use the antigravity sdk skill (install it if it’s not there).
Plan your work after exploring and trying out the existing code.
Look at the output of a flight session and download and read the relevant specs of relevant subsystems.
Connect the antigravity SDK using Agent Platform.
Use Gemini 3.8 flash in the global region.