Power agent hubs or custom harnesses with the Antigravity SDK in one toolkit

Enterprise teams building custom agent hubs often need predictable runtimes, comprehensive logging, and sandboxed execution. The Antigravity SDK provides the same runtime engine used in Antigravity 2.0 and the Antigravity CLI, offering declarative safety policies, real-time telemetry, and stateful multi-turn persistence directly within applications. Updates to the core runtime automatically propagate to SDK-powered agents, ensuring consistent performance and governance across deployments.

A multi-agent control plane powered by the Antigravity SDK enables operators to monitor and manage LLM workloads in real time. It tracks agent reasoning, tool usage, and state storage, providing visibility into actions such as model interactions, tool calls, and skill execution. The SDK’s observability layer streams telemetry via WebSockets to dashboards, allowing teams to track agent progress without manual intervention.

The SDK integrates with existing workflows by resolving skills from filesystem paths containing SKILL.md bundles, which dynamically enhance agent capabilities without external registries. Built-in tools for file and workspace operations execute strictly within configured directories under declarative safety policies, eliminating the need for custom filesystem wrappers and reducing operational overhead.

State persistence is handled through declarative configuration, with the save_dir parameter establishing a root directory where full turn trajectories, tool receipts, and artifacts are stored under traj-. Lifecycle hooks enable real-time interception of agent transitions, allowing dashboards to stream status updates, validate tool calls, and enforce human-in-the-loop approvals before execution.