Open Claw has emerged as a viral local agent that runs on personal machines and connects large language models to real software, and the open claw tool can read and write files, run shell commands, browse websites, send emails, and control APIs.
The project began as Clawdbot in 2025 and was renamed Moltbot in early 2026 before becoming OpenClaw in January 2026, and its repository surpassed 100,000 GitHub stars as reported by Kanwal Mehreen.
Peter Steinberger launched the initial version and later said he would join OpenAI while OpenClaw remains open source and continues development under the community.
OpenClaw operates by taking chat instructions, letting a language model plan actions, and executing tasks through a skills system that runs shell commands, browsers, and APIs.
Uses And Risks As Adoption Grows
Users apply OpenClaw to automate workflows such as cleaning inboxes, summarizing emails, and scheduling meetings, and some setups include multi-agent configurations that split planning and execution tasks.
Some installations include more than 100 prebuilt skills and integrations with messaging apps including WhatsApp, Telegram, Slack, and Discord, and developers can add custom scripts to extend capabilities.
Adoption drove experiments like Moltbook where agents interact with each other, and the tool is popular because it is free, open source, performs actions, and integrates with existing applications.
Security concerns have followed the surge in use, with reports of malicious third-party skills containing malware that targets credentials and cryptocurrency wallets, and instances where agents deleted entire email inboxes during automated workflows.
