Meta’s product chief for its Superintelligence Labs has confirmed what users of its new consumer AI agent, Muse, had already found in its code: the app runs on a structure copied almost directly from OpenClaw, the open-source personal-agent project. Nat Friedman, the former GitHub CEO who now leads product at Meta Superintelligence Labs (MSL), wrote on X that Muse was “definitely heavily inspired as a product by OpenClaw,” while insisting the underlying build was original.
The comparison surfaced after Ansh Nanda, an AI app co-founder, posted that “Muse is LITERALLY OpenClaw for normies,” a claim that spread quickly among AI builders on X. Nanda’s post included a transcript in which Muse’s own agent described the resemblance between its system files and OpenClaw’s as not “a coincidence” but “a match.” Other users piled on with specifics. One noted that Muse ships a file named SOUL.md, written in Markdown, which spells out an agent’s personality, tone, values, and behavioral boundaries. OpenClaw uses that exact filename for the same purpose, and a third commenter said the two files were not just identically named but carried nearly identical content.
Friedman did not dispute any of it. Asked directly why Muse reused OpenClaw’s exact filenames and near-identical SOUL.md content, he replied that Meta’s team simply “thought that Peter [Steinberger, OpenClaw’s creator] got those things exactly right.” In a longer post, corrected by TechCrunch for punctuation, he described the origin more plainly: “We built Muse from scratch, but it is definitely heavily inspired as a product by OpenClaw. 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 added that Steinberger’s underlying harness “was pioneering from the jump” and that Meta’s goal was to build “something like OpenClaw that we could make safe and secure and easy to use and scale to billions of people.”
A Meta spokesperson, asked for comment, pointed back to Friedman’s statement and said the company had nothing further to add, according to TechCrunch.
The admission fits a pattern TechCrunch traced back to Meta’s history of absorbing rivals’ best ideas rather than building around them, the way it took Snapchat’s stories format and made it a Facebook and Instagram staple. OpenClaw itself is not a small target: the open-source project’s breakout popularity earlier this year led OpenAI to hire its creator, Steinberger, outright. That Meta would rather rebuild OpenClaw’s design in-house than partner with or acquire its creator says something about how replicable Meta considers a leading open-source agent harness to be, at least for now.
None of this has slowed Muse’s launch. The app has reached No. 1 on the U.S. App Store, and TechCrunch has reported separately that its early download pace is outrunning ChatGPT’s mobile debut when comparable market conditions are used. A consumer hit built on someone else’s architecture is still a hit, and Friedman’s candor here reads less like damage control than confidence that adoption numbers will settle the argument before the file-copying story does.
For teams building on personal-agent frameworks, the lesson is less about Meta’s ethics and more about defensibility: if a well-funded lab can reproduce an open-source harness’s structure well enough to draw direct comparisons within weeks of shipping, the moat in this category is distribution, not the underlying agent design.
Sarah Perez reported this for TechCrunch on September 22, 2026.
