AIAny
AI Agent2026
Icon for item

AG Kit

Installs and manages a complete .agents workspace for Google Antigravity, providing agent engineering primitives—rules, skills, specialist agents, workflows, persistent memory, MCP guidance, orchestration, and a built-in PreToolUse safety hook for auditable, rollback-friendly agent workflows.

Introduction

Most agent toolkits promise orchestration and helpers; AG Kit makes those guarantees reproducible and native by targeting the Antigravity runtime directly. That placement trades portability for tighter safety, packaging, and CI contracts—so teams running production agents inside Antigravity get stronger runtime assertions and simpler compliance.

What Sets It Apart
  • Antigravity-native workspace contract: installs a managed .agents/ workspace with manifest, SemVer contracts for agents/skills/workflows, and a SHA-256 inventory for packaged plugins—so updates are merge-aware and recoverable.
  • Built-in safety gate: a narrow PreToolUse hook blocks high-risk, destructive tool calls (root fs deletion, drive formatting, raw-disk overwrite) while allowing normal developer cleanup, making destructive-command checks automatic in the runtime.
  • Opinionated orchestration and MCP guidance: includes 20 role definitions, 47 skills, 13 workflows, rules, and explicit MCP sync helpers and backup policies—providing a ready set of specialist routing and persistent-memory conventions for multi-agent flows.
  • Production readiness checklist and validation: CI-ready validators, Antigravity Doctor checks, dependency review and a documented production smoke test reduce operational surprises when promoting an agent suite to production.
Who it's for and trade-offs

Great fit if you operate or plan to operate an agent-driven codebase inside a trusted Google Antigravity workspace and need reproducible update/rollback semantics, built-in safety gating, and an opinionated workspace layout. It accelerates teams that value auditability, manifest-driven packaging, and runtime-enforced safety policies.

Look elsewhere if you need a runtime-agnostic agent framework (AG Kit is explicitly Antigravity-native), if you cannot run Node.js 22 / Python 3.10 in your toolchain, or if you require lightweight toy agents rather than a managed, production-focused workspace. The project intentionally narrows the safety hook surface and does not replace platform permissions or sandboxing.

Information

  • Websitegithub.com
  • OrganizationsUnikorn, Google Antigravity
  • AuthorsVudovn
  • Published date2026/01/14

Categories

More Items

Hugging Face
AI Model2026

Multimodal agentic model for long-horizon computer and browser tasks, with visual self-correction and function-calling. The Pro variant is a 397B Mixture-of-Experts (≈17B active) model with a 262,144-token context window, Docker deployment recipes, and weights currently marked “coming soon.”

Hugging Face
AI Model2026

A multimodal, agentic LLM optimized for long‑horizon, visually grounded workflows — capable of operating browsers and terminals and autonomously executing and testing code. Open‑source weights are available and the family ships in mini, Pro and Max variants for different compute/quality tradeoffs.

Hugging Face

Provides ~483K agent instruction‑tuning trajectories for supervised fine‑tuning, including tool calls, environment feedback, errors/retries and verification across search, code, office and general agent workflows; static snapshots for SFT and mix‑ratio studies.