Two weeks after Meta launched Muse and watched it climb to No. 1 on the U.S. App Store, the company's product chief has conceded what early adopters had been saying for days: the hit personal agent looks like OpenClaw because Meta modeled it on OpenClaw. Nat Friedman, head of product at Meta Superintelligence Labs (MSL), wrote on X that Muse was built from scratch but is "definitely heavily inspired as a product by OpenClaw," the open source agent whose creator, Peter Steinberger, was hired by OpenAI in February.
The admission, first reported by TechCrunch's Sarah Perez on September 22, came after users found that Muse shares identical filenames for several workspace files with OpenClaw, including SOUL.md, the plain-text Markdown file that sets an agent's personality, tone, values, behavioral boundaries and expertise. Others went further and compared the contents, finding the two SOUL.md files nearly identical.
How the comparison unfolded
The pile-on started with a viral X thread from Ansh Nanda, an AI app co-founder, who declared that "Muse is LITERALLY OpenClaw for normies." Nanda posted a conversation in which Muse itself described the overlap between its system files and OpenClaw's as not a coincidence but "a match." Within hours, other users were posting side-by-side comparisons of the agents' workspace directories.
Friedman, the former GitHub CEO who joined Meta last year alongside chief AI officer Alexandr Wang, did not push back. "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 wrote, in a post TechCrunch lightly edited for punctuation.
He was effusive about Steinberger. "@steipete 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," Friedman wrote.
When Nanda pressed him on why a from-scratch product would reuse OpenClaw's exact workspace filenames and near-identical SOUL.md content, Friedman again declined to dispute the premise, replying that "we thought that Peter got those things exactly right." According to The Next Web, the two replies went up minutes apart on Monday evening as personal posts rather than a corporate statement. Asked for comment, a Meta spokesperson pointed TechCrunch to Friedman's post and said the company had nothing to add.
A hit with a borrowed blueprint
The candor lands at a moment when Muse is the biggest consumer AI story of the month. Meta launched the agent on September 8, and Stocktwits, via Yahoo Finance, reported that it logged more than 902,000 downloads in its first six days, outpacing the original Meta AI app. The Information reported, citing internal Meta data, that more than 500,000 people tried Muse in roughly its first week, including over 250,000 daily users who submitted more than two million prompts. Meta shares rose 11.4 percent on Monday, the stock's best single-day gain since April 2025.
Legally, there is little for Meta to answer for. Steinberger released OpenClaw under an open source license, which, as The Next Web noted, turns the file overlap into a product question rather than a legal one. OpenClaw, which launched as Clawdbot and was briefly renamed Moltbot, broke out among developers in January and later spread widely, including in China, where major platforms built hosted versions for their own ecosystems.
Why It Matters
Friedman's answer is an unusually frank acknowledgment of how the agent race is actually being run. The hard, differentiating work in personal agents increasingly sits not in the model but in the harness around it: the instruction files, tool definitions and control loop that decide what an agent does next. OpenClaw proved which harness conventions users liked, and Meta, by its own product chief's account, adopted them wholesale rather than reinventing them. That is permissible under open source norms, and Friedman, who ran GitHub, knows those norms well. But it also underlines that a single independent developer set the template that a trillion-dollar company is now scaling to billions.
It also fits a familiar Meta pattern. TechCrunch pointed to the company's long history of adopting rivals' best ideas, most notoriously Snapchat's Stories format, which Instagram cloned in 2016, and which Mark Zuckerberg acknowledged at a 2020 antitrust hearing. The difference this time is that the source is open code, not a competitor's closed product, and the original author now works for OpenAI, Meta's most direct rival in consumer AI.
What Meta adds, and what OpenClaw lacks, is distribution and control. Meta reviews and approves the connectors that plug outside services into Muse, and it has already lined up payment integrations with Shopify and Stripe. The open project has no gatekeeper; Meta's version has one, and that gatekeeper is Meta.
What to Watch
The first thing to watch is whether Steinberger or OpenAI responds. Friedman's praise was generous, but OpenAI now employs the person whose design choices underpin Meta's breakout product, and it has every incentive to frame Muse as derivative. Second, expect scrutiny of OpenClaw's license terms and attribution requirements as developers dig further into Muse's workspace files. Third, the safety pitch Friedman made, building "something like OpenClaw" that is "safe and secure," will be tested quickly, as Muse asks users for access to email, calendars, messaging accounts and macOS permissions at a scale OpenClaw never reached. If Meta can deliver that security while keeping the OpenClaw feel, the borrowed blueprint will look like a shrewd bet. If it cannot, the comparison will cut the other way.
“We built Muse from scratch, but it is definitely heavily inspired as a product by OpenClaw.”— Nat Friedman, Head of Product, Meta Superintelligence Labs