OpenClaw
Open-source personal AI agent that runs on your own computer
Profile
- First released
- November 2025 (started as 'WhatsApp Relay' and 'Clawd')
- Created by
- Peter Steinberger (Austrian developer) and the community
- Maintained by
- OpenClaw Foundation (US 501(c)(3) non-profit, launched July 2026)
- License
- MIT License
- Type
- Open-source personal AI agent (runs on the user's own devices; no paid tier)
OpenClaw is an open-source (its code is published so that anyone can read, change and reuse it) personal AI agent (an AI that carries out tasks people give it, such as managing a calendar) that runs on the user's own computer. It began as a personal project that the Austrian developer Peter Steinberger built over a weekend in November 2025. It was first called 'WhatsApp Relay' and then 'Clawd', a pun combining Claude with a lobster's claw. It was renamed 'Moltbot' on 27 January 2026 at Anthropic's request, and then 'OpenClaw' on 29 January.
Users hand it tasks through messaging apps such as WhatsApp, Telegram and Slack. Its memory and login details are stored on the user's own device, and users choose an AI model such as Claude, Codex or a model running on their own computer. Within two months of release it had passed 100,000 stars (a mark of interest) on GitHub, a code-sharing site. But because it operates the computer directly, security problems also emerged. In February 2026 a flaw was made public that could send an authentication token (a value that works like a login key) to another server without asking (rated 8.8 out of 10 for severity); it was fixed in version 2026.1.29. In March, Chinese authorities restricted the use of OpenClaw on office computers at state-owned enterprises and government agencies.
In February 2026 Steinberger announced that he was joining OpenAI and said that OpenClaw would move to a foundation and remain an open, independent project. On 8 July 2026 the OpenClaw Foundation, a US non-profit, was launched and said it would keep OpenClaw under the MIT License (a set of usage rules that lets anyone freely use and modify it). Steinberger still makes the technical decisions. OpenAI, Amazon and others support the foundation, but they do not own the project. In September Microsoft introduced Autopilot, a personal agent built on OpenClaw, and on 29 September the foundation released OpenClaw Enterprise for companies. As of October 2026 OpenClaw has no paid tier. OpenClaw shows both the convenience and the security risks of AI agents that operate computers on people's behalf.
Sources for this profile (11) · as of Oct 31, 2026
- openclaw/openclaw · GitHub (OpenClaw Foundation)
- Introducing OpenClaw · OpenClaw Blog
- Clawdbot creator says Anthropic 'forced' him to rename the viral AI agent · Business Insider
- OpenClaw's AI assistants are now building their own social network · TechCrunch
- OpenClaw creator Peter Steinberger joins OpenAI · TechCrunch
- steipete.me – post announcing he is joining OpenAI and moving OpenClaw to a foundation (2026) · Peter Steinberger (steipete.me)
- CVE Record: CVE-2026-25253 · CVE Program (MITRE)
- China Moves to Limit Use of OpenClaw AI at Banks, Government Agencies · Bloomberg
- Introducing the OpenClaw Foundation · OpenClaw Blog
- Microsoft Autopilot is built on OpenClaw. The contributions go both ways. · OpenClaw Blog
- OpenClaw Enterprise - The Open Agent Platform · OpenClaw Blog
AIMade by AIThis profile was written by AI (Claude) from the 11 sources above (official pages, filings, news reports).How it's made
Records 12
Newest first. Open one to see it with its source.
As skill and plugin marketplaces grow, the supply chain becomes an attack surface. A structure that runs vetting through benchmarks from multiple participants rather than a single vendor follows the path package reposito
A path has opened for agent runtimes to move from vendor products to foundation-owned standard infrastructure. The key is defining the harness, model and sandbox as swappable slots: if this separation takes hold, the cos
If agents are treated as always-on operational infrastructure, a failed update is not a problem with one feature but the loss of the means of recovery itself. A case of the principle of zero-downtime deployment coming do
An agent project that had kept a rapid-release culture moves to a system of LTS releases and maturity disclosure
Recovery safeguards on the level of consumer software come to an agent runtime
A structure in which competing labs jointly sponsor a neutral agent standard
Reveals that the agent runtime's permission boundaries were imprecise on every front
A skill distribution standard was proposed as a vendor-led open spec, beyond a single project
A public admission by the maintainers that agent runtimes cannot be defended with language-level guardrails
An early case of bringing app-store-style pre-screening to the distribution of agent skills
The first major incident to shatter the assumption that local agents are safe. Origin validation has since become a baseline requirement
The 3rd renaming in 2 months. A turning point that separates the project's identity from a personal brand