Autonomous AI agent — 335K stars, powerful but security-challenged
OpenClaw is the fastest-growing open-source AI agent in history, hitting 335,000+ GitHub stars in just 60 days — breaking React's decade-old record. Unlike Claude Code or GitHub Copilot (which are code editors), OpenClaw is an autonomous task execution agent that operates across messaging platforms, browsers, and APIs. It supports 50+ integrations and 12 LLM providers. However, OpenClaw faces critical security challenges: 9+ CVEs discovered in its first 2 months, a malicious skill registry (20% of submissions infected), and 30,000+ exposed instances found on the open internet. Created by Austrian developer Peter Steinberger, who recently joined OpenAI.
Not a code editor — an autonomous agent that executes complex multi-step tasks across platforms, browsers, APIs, and messaging systems
Connects to Slack, Discord, email, browsers, file systems, databases, and CI/CD pipelines for end-to-end automation
Supports Claude, GPT, Gemini, Llama, Mistral, DeepSeek, and any OpenAI-compatible API including local models
Extensible via ClawHub skill registry — community-contributed automations. WARNING: 20% of submissions found to contain malicious code
$0
Bring your own API keys. Security is your responsibility
Via NVIDIA
NVIDIA NemoClaw wraps OpenClaw for enterprise use
Runs as CLI, Discord bot, Slack app, or web service — operates autonomously with minimal supervision
MIT licensed, 335K+ GitHub stars, community governed. Founder Peter Steinberger recently joined OpenAI (Feb 2026)
OpenClaw's latest release, version 1.5.0, brings significant solver enhancements, improved parallel processing, and better AMR support for CFD simulations.
OpenClaw v2.1.0 is now available, introducing an advanced anomaly detection model with improved accuracy.
CRITICAL: 9+ CVEs, malicious skill registry, 30K exposed instances. What you need to know.