Full Stack HQ: A Permission-First AI Engineering Stack for Claude Code, Antigravity, and Codex
AI coding agents can move quickly. The engineering challenge is keeping intent, approval, and verification visible while they do. That is the problem Full Stack HQ is designed to address: a shared, permission-first engineering workflow that can travel across multiple AI coding hosts. The project supports Claude Code, Google Antigravity IDE, and OpenAI Codex from one tool-agnostic source of truth.…
Full Stack HQ is an AI engineering stack designed to provide a permission-first workflow for AI coding agents, such as Claude Code, Google Antigravity IDE, and OpenAI Codex. The primary challenge in AI coding is keeping intent, approval, and verification visible while the agent is working. Full Stack HQ addresses this by creating a shared, permission-first engineering workflow that works across multiple AI coding hosts.
The core problem with AI coding agents is that most users configure them once (or never) and let them run without oversight. As a result, the agents make assumptions about what the user wants, take irreversible actions without asking, mix planning and execution in the same step, and lack consistent code style or architectural awareness. This unpredictable combination can be the worst in software development.
The solution proposed by Full Stack HQ is a permission-first workflow where the agent plans, shows what it intends to do, and requests explicit approval before execution. Users can add user authentication with JWT, and the agent will present its plan with phases that require approval before proceeding. For example, adding user authentication with JWT would have the agent requesting approval for each phase, such as creating an auth module and JWT strategy, adding guards to protected routes, and implementing refresh token rotation.
Full Stack HQ consists of various components, including a shared rule core, host-neutral engineering policy, host adapters for Claude Code, Antigravity, and Codex, generated host instruction files, specialist AI personas, skills, workflow files, and more. Instead of using a single generic agent for everything, Full Stack HQ provides domain experts tailored to specific tasks, such as frontend specialists (React, Next.js, Tailwind), backend specialists (NestJS, Node.js, APIs), database specialists (Prisma, PostgreSQL), architects, system design experts, code reviewers, quality assurance agents, test engineers, security auditors, performance optimizers, and DevOps engineers.
To install Full Stack HQ, users can run a script from the cloned repository, specifying which host they want to install on (Claude Code, Antigravity, or OpenAI Codex). The installer performs pre-flight checks and installs the selected host files, keeping the Antigravity legacy form and rendering each host as a skill for Claude Code, Codex, and modern Antigravity.
The shared rule-core philosophy of Full Stack HQ emphasizes the importance of separating planning and execution, role-based reasoning, enforcing explicit code style, and including a security checklist before every commit. These principles help the AI agents behave more like senior engineers who think, propose a plan, get sign-off, and execute one reversible step at a time.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
