HIVE gives you 11 specialized squads — each with its own persona, memory, and skills — orchestrated by an AI Chief of Staff.
The real leverage is in Phase 3 — but building that infrastructure from scratch takes months. HIVE is that architecture, ready to clone.
ChatGPT, prompts, personal productivity. Useful, but individual and disposable.
APIs, automations, one-off workflows. Still reactive, still stitched together.
Systems with context, memory, and specialized agents that work like a real team.
HIVE is a folder structure and a set of conventions. No framework lock-in — just Claude Code, markdown, and Python.
Each squad is a Claude Code agent with its own CLAUDE.md, STATE.md, decisions log, and reusable skills.
Context, memory, hooks, and skills — the building blocks that make HIVE feel like a team instead of a chatbot.
Each squad's CLAUDE.md defines its persona, scope, and rules. Claude Code loads it when you navigate to the squad folder — no copy-pasting prompts between sessions.
STATE.md holds live operational state in three layers: L1 (right now), L2 (in progress), L3 (backlog). Decisions and gotchas persist indefinitely.
9 hooks handle session branches, squad routing, state updates, and incident lookups — without you managing any of it manually.
Invoke with /skill-name. Ships with /hive-setup, /open-squad, /status, /close-squad — and you write your own for your company.
Clone, open Claude Code, run onboarding. That's the full install process.
# 1. Clone the repo git clone https://github.com/felipeluissalgueiro/hive.git my-company cd my-company # 2. Install Claude Code (skip if you have it) curl -fsSL https://claude.ai/install.sh | sh # 3. Open and run guided onboarding (~10 min) claude /hive-setup # Start working with a squad /open-squad commercial # See what's happening across all squads /status