--- title: "Frequently Asked Questions" description: "12 common questions about Mantis AI Controller answered with specific numbers, version details, and direct comparisons." --- # Frequently Asked Questions 12 answers about Mantis AI Controller, each starting with a direct one-sentence response. ## 1. What is Mantis AI Controller? It is a safety-first controller layer that wraps Claude Code with live hook routing, fail-closed checks, persistent memory, browser testing, and multi-agent orchestration. Think of it as an operating system for AI-assisted coding -- it does not replace Claude Code, it governs how Claude Code operates on your codebase. ## 2. How does the safety gate system work? Every tool call Claude Code makes passes through lifecycle hooks and routed controller batches before execution. The live Claude audit shows 6 lifecycle events, 35 hook entries, 25 controller-owned entries, 20 registered handlers, 15 routed batch aliases, and 37 explicit main-path checks in `pretooluse.py`. If any check fails or crashes, all edits stop. ## 3. Is it really free? What's the catch? Yes, it is completely free during the beta period with no credit card, no account creation, and no usage tracking. The value stack of all components totals $2,611/month based on comparable commercial tooling. The only requirement is a Claude Code subscription from Anthropic ($20/month for Pro or $200/month for Max), which provides the underlying AI model that Mantis controls. ## 4. What platforms does it support? Windows 10+ with WSL2 (version 2.0.0 installer) and native Linux (coming soon). The Windows installer is an Inno Setup executable. The Linux installer is a self-contained shell script. Docker Desktop is optional and only needed for cross-terminal coordination between multiple Claude sessions. ## 5. How is this different from using Claude Code without it? Without Mantis, the AI edits code the moment it has a guess -- no research requirement, no plan approval, no test enforcement, and no memory between sessions. With Mantis, tool calls route through memory, research, planning, approval, testing, backup, deployment, and verification checks. Failed approaches are remembered across sessions so mistakes are not repeated. ## 6. What is the memory system? It is a two-tier persistent memory that carries knowledge across sessions using TF-IDF relevance scoring. Core memory (MEMORY.md, approximately 200 lines) loads at every session start. The append-only archive (memory_log.jsonl, 940KB+ and growing) stores 8 types of knowledge including failed approaches, working solutions, and user preferences. Recent learnings score higher via time decay, and matches are injected at 3 tiers (HOT, WARM, COLD) based on relevance. ## 7. Can I bypass the gates when I need to? Yes, say "maintenance" to temporarily bypass all gates for infrastructure work. Maintenance mode auto-expires after 1 hour as a safety net. Backup gates (project snapshot and per-file checkpoint) continue running even in maintenance mode. Say "done" to re-engage all gates immediately. This is the only supported bypass mechanism -- the system detects and blocks 30+ other bypass patterns. ## 8. How does browser testing work? agent-browser uses your real Chrome installation via direct WebSocket CDP connection with no navigator.webdriver flag, making it completely undetectable. It reads page structure via accessibility tree snapshots using approximately 200 tokens per page compared to 13,700 for Playwright MCP -- a 98.5% reduction. 140+ commands cover navigation, interaction, network interception, device emulation, video recording, and visual regression testing. Parallel sessions via AGENT_BROWSER_SESSION environment variable enable 100s of concurrent agents without race conditions. Ships with a Claude Code skill that researches UI structure, plans interaction paths, and verifies results. Apache-2.0 open source. ## 9. What skills are bundled? Mantis ships the controller workflow skill plus the skills the system depends on: research, planning, review, testing, browser automation, frontend design, humanizer/copywriting, media, and specialist workflow skills. The point is not just a library count; the installed skill tells Claude or Codex when to use the deeper Mantis workflows. ## 10. Does it comply with the EU AI Act? The hook audit trail provides governance tooling ahead of the August 2, 2026 enforcement date. AI decisions pass through documented checkpoints with timestamps, confidence scores, approvals, and backups. This does not constitute legal compliance certification, but it provides technical infrastructure that compliance frameworks often require. ## 11. How long does installation take? The private install is command-line based. The generated setup command installs the `mantis` launcher, controller runtime, skills, update checks, and backend selection. On start, `mantis` asks whether to run Claude or Codex. ## 12. Is it open source? Some bundled skills come from open-source originals or public tooling, but the controller infrastructure, memory behavior, hook routing, and packaged Mantis runtime are proprietary/private. The underlying Claude Code CLI is developed by Anthropic. ## Related - [Getting Started](https://erebora.org/mantis/docs/getting-started) -- Installation walkthrough - [Safety Gates](https://erebora.org/mantis/docs/safety-gates) -- Full gate pipeline reference - [Memory System](https://erebora.org/mantis/docs/memory-system) -- How cross-session memory works - [Browser Testing](https://erebora.org/mantis/docs/browser-testing) -- CDP automation details - [Skills & Agents](https://erebora.org/mantis/docs/skills-and-agents) -- Skills, slash-command workflows, and agents - [Live Documentation](https://erebora.org/mantis/) -- Full product page