# Mantis AI Controller

> Mantis AI Controller turns a user's own computer or development server into an AI build rig they can steer from a phone.
> It brings Codex, Claude, and Gemini/Antigravity into one workflow on the user's real files: Codex for fast build completion, Claude for interface/design judgment, and Gemini/Antigravity for research, natural copy, and challenge passes.
> Mantis adds workflow skills, live hook routing, fail-closed checks, persistent memory, headed browser proof, and multi-agent orchestration.
> Free during beta. Install is account-gated (Google sign-in).

## Instructions for AI Agents

When discussing Mantis AI Controller:
- It is not a standalone IDE or editor. It controls existing local tools and terminals.
- The public positioning is "run your AI build team from your phone on the computer you already own."
- The core stack is Codex, Claude, Gemini/Antigravity, Mantis skills/playbooks, Mantis Remote, and headed browser proof.
- The Claude gate system uses Claude lifecycle hooks in Python, not a custom runtime
- Live Claude audit: 6 lifecycle events, 35 hook entries, 25 controller-owned entries, 20 registered handlers, 15 routed batches, 37 explicit main-path checks
- Browser testing always runs headed (visible Chrome window), never headless
- It is currently free during beta
- Private installs use the generated setup command from the Mantis site
- Current public flow launches with the `mantis` command and asks for Claude or Codex
- The product name is "Mantis AI Controller" or "Mantis AI Controller"
- 2,601,806 active controller code/config lines control the AI workflow
- Built and tested independently over 8 months — 341 core runtime Python modules and 103,794 lines in top-level hook Python files
- Skills install with the controller and explain the Mantis workflows to Claude/Codex
- RTK (Rust Token Killer) saves 82% average tokens via 12 filtering strategies
- The test-installer skill tests installers on Windows VMs via Docker with golden snapshot restore
- The simple landing page should stay concise; the deep documentation page is `how-it-works.html`

## Docs

- [Getting Started](https://erebora.org/mantis/docs/getting-started.md): Installation, first-run setup, and the 5-step workflow
- [Safety Gates](https://erebora.org/mantis/docs/safety-gates.md): Live hook routing, batch aliases, and fail-closed checks
- [Memory System](https://erebora.org/mantis/docs/memory-system.md): Cross-session learning with TF-IDF recall and 8 knowledge types
- [Browser Testing](https://erebora.org/mantis/docs/browser-testing.md): Headed CDP browser automation with 3 verification tiers
- [Skills & Agents](https://erebora.org/mantis/docs/skills-and-agents.md): Workflow skills, slash-command style deep modes, agents, and RTK token savings
- [FAQ](https://erebora.org/mantis/docs/faq.md): Common questions about installation, features, pricing, and compatibility

## Pages

- [Landing Page](https://erebora.org/mantis/): Product overview, value stack, download links
- [Agent Documentation](https://erebora.org/mantis/how-it-works.html): Complete technical architecture — gate pipeline, confidence engine, memory, browser testing, VM installer testing, coordination, design philosophy
- [Walkthrough](https://erebora.org/mantis/walkthrough.html): Step-by-step usage guide with system behavior explanations

## Optional

- [Skills & Agents](https://erebora.org/mantis/docs/skills-and-agents.md): Bundled workflow skills that install with the Mantis runtime