Download
NullClaw is the smallest fully autonomous AI assistant infrastructure, built entirely in Zig as a single static binary with zero runtime dependencies. At just 678 KB with ~1 MB peak RAM usage, it boots in under 2 milliseconds and runs on virtually any hardware, including low-cost ARM boards. Despite its size, it delivers a complete AI stack with 22+ model providers, 18+ communication channels, integrated tools, hybrid memory, and sandboxed runtime support. Its architecture is fully modular, using vtable interfaces that allow providers, channels, tools, memory backends, and runtimes to be swapped without code changes. NullClaw is secure by design, enforcing pairing-based authentication, strict sandboxing, encrypted secrets, resource limits, and workspace scoping by default. Designed for portability and independence, it supports OpenAI-compatible APIs, multiple tunnels, hardware peripherals, and edge deployments including WASM-based logic.
Features
- Ultra-Lightweight Binary: 678 KB static Zig binary with no VM, framework, or runtime overhead.
- Near-Zero Resource Usage: ~1 MB RAM usage and sub-8 ms startup even on low-frequency edge hardware.
- Extensive Provider & Channel Support: 22+ AI providers and 18+ messaging channels including Telegram, Discord, Slack, Nostr, WhatsApp, and more.
- Hybrid Memory System: SQLite-backed storage combining vector embeddings with FTS5 keyword search and weighted merging.
- Security-First Architecture: Pairing-based authentication, sandbox isolation (Landlock, Firejail, Docker, etc.), encrypted secrets, and audit logging.
- Fully Modular & Swappable: Pluggable providers, tools, runtimes (native, Docker, WASM), tunnels, peripherals, and observability systems via vtable interfaces.
Project Samples
