Index / LightClaw · updated Sep 1, 2026
LightClaw
zofrasca/lightclaw · healthy · rank 36 of 43 by stars
A single-binary Rust reimagining of OpenClaw with near-instant startup and a ~15MB footprint. It targets users who want agentic tooling, memory, and Telegram/Discord integration without a heavy runtime.
Facts
Repository
Runtime
Posture
Community
Security breakdown
Composite 62 / 100 · how these are scored
higher is safer
higher is safer
higher is safer
higher is safer
higher is riskier
Model access
Read from the repository, not written by a model · 35 files examined
2 providers · runs locally · gateway support
Pinned models
claude-opus-4-5 released 2025-11-24 — 9 months old (from the public model catalogue)
Pin last edited 6 months ago
Evidence
Decision
Why choose LightClaw over OpenClaw?
Why choose this
- Single ~15MB binary with no runtime dependencies
- Near-instant startup versus OpenClaw's heavier boot
- Broad platform support including ARMv7 and Raspberry Pi
Tradeoffs
- Much smaller community and ecosystem
- Windows support is unstable
- Fewer integrations and skills than OpenClaw
Best fit
- Users wanting a minimal single-binary OpenClaw alternative
- Raspberry Pi and low-resource ARM deployments
- Telegram/Discord-native personal assistants
Avoid if
- You need Windows as a primary platform
- You require a mature plugin ecosystem
- You want strong sandboxing around shell tools
README and commit evidence are solid for architecture and footprint claims, but there is zero Reddit discussion and minimal independent web coverage, so community sentiment and long-term viability are uncertain.
AI layer reviewed Aug 24, 2026 · how this is written
Star activity
223 stars today
Overview
LightClaw is a Rust-based, single-binary AI assistant inspired by OpenClaw and nanobot, built on the Rig framework. It ships as a ~15MB executable with no Python runtime or virtual environment, targeting near-instant startup and a low memory footprint across Linux (x86_64, ARM64, ARMv7) and macOS, with a less-stable Windows build.
Architecturally, it provides a tool-capable agent with file, shell, web, and scheduling actions, plus Telegram and Discord interfaces via high-performance polling. Its memory system combines short-term session history, periodic summarization, and semantic retrieval over locally stored vectors in SQLite, keeping data privacy-first with no external vector database. It also supports OpenClaw-style SKILL.md skills through an activate_skill mechanism, and recent commits show active work on XDG config support, ARMv7 builds, and shell guard refinements.
Compared to OpenClaw, LightClaw trades ecosystem breadth and maturity for radical simplicity: one binary, one config file, and local-first storage. It is best suited for personal, single-user deployments on desktops or low-power edge devices rather than multi-tenant or heavily integrated setups.