Not a Coincidence: Meta Admits Muse Was 'Heavily Inspired' by OpenClaw

When Meta launched Muse — its consumer AI agent and the centerpiece of its post-superintelligence-lab product push — early adopters noticed something odd. The agent’s internal workspace didn’t just resemble OpenClaw, the wildly popular open-source personal agent framework. It looked the same: the same file names, the same structure, and a SOUL.md personality-configuration file whose content was nearly identical to the one OpenClaw ships with.

For weeks, the community treated it as a running joke: “Muse is literally OpenClaw for normies.” On September 22, Meta stopped pretending. Nat Friedman — former GitHub CEO, now head of product at Meta’s Superintelligence Labs (MSL) — confirmed on X that Muse was “definitely heavily inspired as a product by OpenClaw,” while insisting the software itself was “built from scratch.”

The clarification came in response to a viral post from Ansh Nanda, an AI app co-founder, whose side-by-side comparison of the two agents’ system files set the AI corner of X ablaze. One person flagged that Muse carried a SOUL.md file — the plain-text Markdown configuration that defines an AI agent’s personality, tone, values, and behavioral boundaries, a convention OpenClaw popularized. Another found that the files weren’t merely named the same; their content was almost entirely the same. When pressed on why Muse had copied OpenClaw’s exact workspace file names, Friedman didn’t dispute it. “We thought that Peter got those things exactly right,” he replied, referring to Peter Steinberger, OpenClaw’s creator.

From January obsession to product blueprint

Friedman’s admission came with an origin story. “After I used OpenClaw in January, I bought hundreds of Mac minis for the MSL team, and lots of us fell in love with using OpenClaw (and other personal agents),” he wrote. “Peter is a genius, and his harness was pioneering from the jump. I think a lot of people were inspired by it. Our goal with Muse was to build something like OpenClaw that we could make safe and secure and easy to use and scale to billions of people.”

That last clause is the whole strategy in one sentence. OpenClaw is famously powerful and famously fiddly — a self-hosted harness aimed at developers willing to wrangle API keys, config files, and their own hardware. Meta’s bet is that the interaction paradigm Steinberger pioneered is the correct one, but that reaching a mainstream audience requires packaging it with managed safety, security, and onboarding. Friedman’s hundreds of Mac minis — bought so the entire MSL product team could live inside OpenClaw daily — were the due-diligence step.

The timing matters. Steinberger built OpenClaw essentially alone in about three months, and it became the fastest-growing open-source project in GitHub history — blazing past 200,000 stars and eventually hitting some 250,000 within roughly 60 days of its breakout. OpenAI was impressed enough to recruit Steinberger in February 2026, a hire widely read as a signal that personal agents were the next platform war. Friedman himself joined Meta last year alongside Chief AI Officer Alexandr Wang as part of the company’s talent sweep.

Open source’s most expensive compliment

There is no legal drama here — OpenClaw is open source, and studying it is fair game. But the episode is a textbook example of Meta’s well-worn playbook: identify a promising product, then copy its best features. The company did it most notoriously with Snapchat’s Stories format, which it cloned into Instagram and turned into an industry-wide default.

What’s different this time is the source. OpenClaw isn’t a rival startup’s product; it’s a community project run on volunteered labor and one developer’s obsessively documented usage. Meta copying it is simultaneously the biggest possible validation of open-source agent design and a pointed reminder of where value accrues: Steinberger’s ideas now ship inside a Meta product that recently hit No. 1 on the U.S. App Store — data indicates Muse is outpacing ChatGPT’s launch when platform availability is compared directly — while OpenClaw itself continues as an independent project whose creator now works for OpenAI.

For open-source maintainers, the lesson is double-edged. Permissive licenses mean frontier labs can absorb your architecture, your file conventions, even your documentation patterns, without a phone call. But it also means the conventions a solo developer invented on nights and weekends — a Markdown file named SOUL.md defining an agent’s soul — are now en route to billions of people.

Why the SOUL.md detail matters

Beyond the corporate-strategy narrative, the specificity of the borrowing is striking. OpenClaw’s design insight is that an agent’s persona and boundaries should live in readable, editable text files inside the workspace itself — not buried in a system prompt or a settings panel. Users can open SOUL.md, read exactly what their agent believes it is, and rewrite it. That transparency-as-configuration pattern is what Friedman says Meta “thought Peter got exactly right,” and it is now embedded in one of the world’s most-downloaded apps.

Reached for comment, a Meta spokesperson pointed to Friedman’s statement and said the company had nothing more to add.

As Meta prepares to detail Muse’s next chapter at its Connect conference on September 23–24, the admission reframes the product story: the frontier of consumer AI agents may be fought less over who invents new paradigms and more over who industrializes the best existing ones — wherever they were born.