🛡️ Chitin Shell

Your first secure AI agent in 10 minutes.

npx chitin-shell

No decisions. No configuration files. No security expertise needed.

Answer 5 questions. Get a production-hardened AI agent with contextual trust boundaries.

What It Does

1. Detects your system (OS, CPU, memory, available tools)
2. Asks 5 simple questions (name, agent name, channel, provider, password)
3. Installs OpenClaw and your chosen AI provider
4. Configures your agent with secure defaults
5. Hardens your system (UFW firewall, fail2ban, file permissions)
6. Deploys Chitin Moat — contextual agent permissions
7. Verifies everything works

The Security Problem

Most agent setups are insecure by default because they require users to make security decisions. Every decision is a place to get it wrong.

Chitin Shell eliminates the decisions. Security is structural, not optional.

Feature	Default	User Choice?
UFW firewall	Deny all incoming	No
fail2ban	Active on SSH	No
Gateway binding	Loopback only	No
Trust channels	Sovereign (owner DM), Observer (everything else)	No
Secrets encryption	Restricted file permissions (700)	No
Sub-agent sandbox	Docker isolation	No

Supported Providers

Provider	Cost	Quality	Notes
🆓 Groq	Free	Good	Llama 3.3 70B, fast inference
🧠 Anthropic	Paid	Best	Claude Sonnet/Opus
🌐 OpenAI	Paid	Great	GPT-4o
🏠 Ollama	Free	Varies	Local, private, needs GPU

Supported Channels

• 📱 Telegram (recommended)
• 💬 Discord
• 💬 Signal
• 💻 CLI (no channel needed)

Dry Run

Test without making changes:

npx chitin-shell --dry-run

What's Next?

After bootstrap, level up with The Vesper Blueprint — a comprehensive guide to understanding and customizing your agent's architecture.

Built By

Chitin.xyz — Trust infrastructure for the agent economy.

License

Apache 2.0