The secure, open-source operating system for AI agents
AI agents are powerful but risky. RustyClaw is built with the assumption that agents can't always be trusted.
| Defense Layer | What It Does |
|---|---|
| 🛡️ PromptGuard | Detects prompt injection attacks, system overrides, and jailbreaks |
| 🔐 LeakDetector | Blocks credential exfiltration — API keys, tokens, SSH keys |
| 📦 Sandbox Isolation | Bubblewrap (Linux), Landlock (5.13+), sandbox-exec (macOS) |
| 🌐 SSRF Protection | Blocks requests to private IPs and metadata endpoints |
| 🔑 Encrypted Vault | AES-256 secrets with optional TOTP 2FA |
| Metric | RustyClaw | Node.js Agents | Python Agents |
|---|---|---|---|
| Memory | ~15 MB | ~150 MB | ~100+ MB |
| Startup | <50 ms | ~500 ms | ~1s+ |
| Binary Size | ~8 MB | ~200 MB | N/A |
| Dependencies | 0 (single binary) | node_modules | venv |
File ops, shell, browser automation, web search, memory, scheduling, multi-agent coordination.
Signal, Matrix, Discord, Telegram, Slack, WhatsApp — one agent, every platform.
Spawn, steer, and coordinate sub-agents. Built-in session management and isolation.
Cron jobs, heartbeats, and wake events. Agents that work while you sleep.
Semantic search, file-based persistence, automatic context management.
Built-in Playwright integration for web scraping and automation.