GitHub Details OpenClaw's Rise to 388,000 Stars

GitHub disclosed for the first time that OpenClaw, a personal AI assistant that runs on a user's own device, had reached 388,000 stars, 81,000 forks and more than 80,000 commits as of Aug. 26, in a blog post profiling the project's maintainers. GitHub called it "the fastest-growing project in GitHub history."

Creator Peter Steinberger started OpenClaw as a weekend project in November 2025, according to the post. It has since drawn a team of maintainers spread across several companies rather than one employer.

Who is actually running the project

Maintainer	Role	Affiliation
Peter Steinberger	Creator	Independent
Vincent Koc	Chief Architect	OpenClaw Foundation
Josh Avant	Member of Technical Staff	OpenClaw Foundation
Brad Groux	Chief Executive Officer	Digital Meld
Sally O'Malley	Principal Software Engineer	Red Hat
Josh Lehman	Engineer	Martian Engineering
Val Alexander	Maintainer	OpenCoven

What GitHub's post says about reviewing AI-written code

Steinberger said he has stopped calling contributions pull requests. "I don't even call them pull requests. I call them prompt requests," he said, according to GitHub's post. Koc said contributors have tried to game trust signals by duplicating other people's work: "People would basically duplicate other people's pull requests, to build credibility," he said. Steinberger's standard for reviewing a contribution, per the post, is not who or what wrote it: "Nobody cares if you wrote the code or not, but we care if you actually thought about this feature," he said.

What the post does not mention

An independent security analysis of OpenClaw published in February on flyingpenguin.com, when the project had roughly half its current star count, described a specific architecture the GitHub post does not address: access control enforced only through prompts rather than code, unsigned markdown skill files with no verification step, an agent capable of modifying its own source, and no process isolation or credential vault separating one skill from another. Steinberger did not dispute the substance of that architecture in the piece; asked about the agent's ability to rewrite its own code, he said, "I didn't even plan it. It just happened."

GitHub's Aug. 26 numbers are themselves already dated for a project growing this fast: at the star count flyingpenguin cited in February, roughly half of August's total, OpenClaw was already drawing security scrutiny for exactly the kind of unsigned, self-modifying code path the GitHub profile leaves out. GitHub's post does not say whether any of the four architecture gaps have since been closed, and it does not say who, if anyone, signs off on a skill file before it reaches a user's device.

The project's maintainers are scheduled to appear at GitHub Universe on Oct. 28 and 29, where GitHub's post says more detail on OpenClaw's security work is expected. Whether that includes the gaps flyingpenguin raised, or an updated star count, GitHub has not said.

The pattern of a single-maintainer weekend project outgrowing its own security model before anyone builds one echoes what happened with GNU Radio's browser port earlier this month, and it is the same category of risk CISA has been tracking in open-source AI tooling more broadly this quarter.

Sources: OpenClaw went viral. Meet the maintainers building and securing it., GitHub, Aug. 26, 2026; OpenClaw creator makes strong case against OpenClaw, flyingpenguin.com, Feb. 18, 2026.