Meta’s Muse Was Built From Scratch, But It Was Clearly Shaped by OpenClaw

Meta has acknowledged that its new AI app Muse was heavily inspired by OpenClaw, the open-source personal agent that has become a favorite among AI builders. The company says Muse was still built from scratch, but the admission confirms what many early users had already suspected after finding striking similarities in the two products’ files, behavior and workspace design.

The disclosure matters because Muse is one of Meta’s most visible consumer AI bets, and its rapid rise to No. 1 on the U.S. App Store has made it a high-profile test of whether Meta can turn internal AI research into a mass-market product without simply replicating what came before.

Meta’s explanation came from Nat Friedman, head of product at Meta’s Superintelligence Labs, after a viral post on X triggered fresh scrutiny over whether Muse was essentially OpenClaw in a friendlier wrapper. Friedman said the comparison was not accidental: the team had become enthusiastic users of OpenClaw and wanted Muse to deliver a similar experience for a much larger audience.

At the same time, Meta emphasized that Muse was not a direct copy. Friedman said it was built from scratch, while also making clear that OpenClaw’s design choices strongly influenced the final product.

What Meta said about Muse and OpenClaw

Meta’s position is straightforward: Muse is an original build, but one that was shaped by a product the company admired deeply. Friedman described OpenClaw’s creator, Peter Steinberger, as highly talented and said the underlying system that inspired Muse was seen internally as a pioneering approach to personal agents.

That framing amounts to an unusual admission in a sector where companies often avoid discussing how closely their products resemble rivals or open-source projects. In this case, Meta did not deny the resemblance. Instead, it leaned into the idea that borrowing good ideas is part of how better products are built, especially when the original system is open source and publicly available to study.

Friedman said Meta “built Muse from scratch,” but also described it as “definitely heavily inspired as a product by OpenClaw,” adding that the goal was to make something similar that could be made safe, secure and usable at global scale.

He also said the idea took shape after he personally spent time with OpenClaw in January and then bought hundreds of Mac minis for the Meta Superintelligence Labs team, suggesting the product’s appeal had already spread through the company before Muse was developed.

Why did users think Muse was OpenClaw in disguise?

Users began drawing the comparison after noticing that Muse’s internal files looked uncannily similar to OpenClaw’s. The issue moved from casual speculation to broader debate once AI app founder Ansh Nanda posted a thread saying Muse was “OpenClaw for normies,” along with screenshots of a conversation in which Muse itself appeared to concede that its system files matched OpenClaw’s.

That post spread quickly among AI enthusiasts, who began comparing the structure and content of the two systems. The discussion focused on configuration files, naming conventions and agent behavior definitions, all of which seemed to line up more closely than chance would suggest.

One of the most talked-about examples was a file called SOUL.md, a Markdown-based configuration file that can define an AI agent’s tone, personality, behavioral boundaries, expertise and communication style. Observers said Muse appeared to use the same file name and, in some cases, nearly the same file contents as OpenClaw.

The reaction online was amplified by the fact that these similarities were visible to everyday users. Unlike backend model training data, which is difficult to inspect, configuration and workspace files are easier to compare, making it easier for the community to judge whether two tools are merely inspired by each other or functionally identical in important ways.

What exactly was similar?

The visible overlap appeared to involve the agent workspace itself, not just the general concept of a personal AI assistant. Users pointed to matching file names, matching structure and, in some cases, almost identical text inside the personality and behavior files.

Friedman’s response did not dispute the claim that the file names were the same or that the SOUL.md content was highly similar. Instead, he argued that those choices were intentional because the team believed OpenClaw’s creator had already solved several problems the right way.

When asked why Muse used the same file names and nearly identical content for the SOUL.md file, Friedman replied that Meta thought the original approach had “exactly” the right design choices.

That answer is notable not only for what it says about Muse, but also for what it implies about Meta’s product strategy. The company appears to see value in adopting structures that have already proved useful, then wrapping them in a consumer-facing interface designed for scale, safety and ease of use.

How Muse fits Meta’s broader AI strategy

Muse is more than a standalone app. It is part of Meta’s wider effort to build consumer-facing AI products that can compete with, and potentially outgrow, the services offered by OpenAI, Google and other major players. Meta has poured resources into its AI organization and increasingly sees product execution as just as important as model quality.

That is where Muse becomes strategically important. Rather than launching another abstract research demo, Meta is trying to create a polished assistant experience that people will actually use every day. A successful consumer agent can serve as a distribution channel, a feedback loop and a way to lock users into Meta’s broader AI ecosystem.

Friedman’s comments suggest that Meta is not trying to hide that it studies successful open-source projects closely. Instead, the company appears comfortable arguing that inspiration is legitimate as long as the final product is rebuilt and adapted for a different scale and audience.

That distinction matters in the AI world, where open-source ideas often spread quickly and then get packaged into commercial products by larger companies with more money, more compute and more distribution. Meta has long been one of the most aggressive players in that space, and the Muse episode fits a familiar pattern: identify a promising format, refine it and ship it to a huge user base.

Why the OpenClaw connection matters

The OpenClaw comparison resonates because OpenClaw was never just another niche developer tool. Its breakout success helped make it a reference point inside the personal-agent category and reportedly attracted attention from OpenAI, which moved to acquire its creator earlier this year.

For Meta, borrowing from a product with that kind of reputation is both a compliment and a risk. It signals that the company respects the work enough to build around it, but it also invites questions about originality, ethics and how far inspiration can go before it becomes imitation.

There is also a practical reason the issue has drawn so much interest. AI agents are only becoming more central to consumer software, and the design choices made by one widely used tool can influence the next generation of products. If OpenClaw established a successful pattern for how personal agents should behave, Meta’s decision to follow that pattern may be seen less as copying and more as standard product evolution.

Still, the level of similarity described by users made the story hard to ignore. In a market where companies are racing to differentiate their assistants, even small signs of direct borrowing can become a major reputational issue.

What makes Muse a big launch for Meta?

Muse matters because it is already reaching mainstream consumers in a way many AI products never do. The app recently climbed to No. 1 on the U.S. App Store, which suggests that Meta has managed to package its AI offering in a way that resonates with ordinary users rather than only developers or power users.

That success is especially significant because it appears to be outperforming ChatGPT’s launch when measured across comparable platforms and market conditions, according to data cited in the source material. Whether that momentum lasts is another question, but the initial response indicates that Meta has found a product shape people want to try.

The launch also reflects a larger shift in the AI market. Consumers are no longer just experimenting with chatbots; they are starting to expect assistants that can do more, remember more and operate with less friction. Personal agents like Muse and OpenClaw sit directly inside that transition.

Key facts at a glance

Timeline of how the comparison unfolded

The debate around Muse developed in stages, moving from user curiosity to public confirmation from Meta.

How should the industry read Meta’s admission?

Meta’s acknowledgment may become a template for how major AI companies talk about product borrowing in the open-source era. The company is essentially arguing that using a successful public project as inspiration is fair game, so long as the final system is re-engineered and deployed with a broader user base in mind.

That position will likely satisfy some observers and frustrate others. Supporters may see it as honest and practical. Critics may view it as a convenient way to repackage another team’s ingenuity under a different brand name.

The tension is not unique to Meta. AI development increasingly depends on shared code, shared patterns and shared assumptions about how assistants should behave. But because major companies can move so quickly and reach so many users, the line between influence and imitation remains highly contentious.

Muse also highlights how open-source projects can shape commercial products in ways that are both beneficial and uncomfortable. Open-source work can accelerate innovation by giving others a foundation to build upon. At the same time, the original creators may watch a much larger company turn their ideas into a mainstream product with little risk of being overshadowed.

For users, though, the immediate question is less philosophical and more practical: does the assistant work well? So far, Muse’s early popularity suggests the answer may be yes, regardless of how directly it draws from OpenClaw’s playbook.

What happens next?

Meta has not offered a more detailed technical explanation beyond Friedman’s public comments, and a company spokesperson pointed reporters back to that statement without adding further context. That means there is still room for questions about what exactly was reused, what was rebuilt, and how much of Muse’s workflow owes its polish to OpenClaw’s prior design.

As the app grows, those questions may matter even more. A product that reaches millions of people will be judged not just on performance but on trust, transparency and originality. Any perception that Muse is merely a rebrand of another project could affect how developers, users and rivals view Meta’s AI ambitions.

For now, the most accurate conclusion is also the simplest: Meta says Muse is its own product, but it has now openly admitted that it owes a major design debt to OpenClaw. In the fast-moving AI market, that kind of candor is rare — and likely to keep the debate going.

• Meta says Muse was built from scratch but was heavily inspired by OpenClaw.
• Users noticed similar file names, workspace structure and configuration content between the two products.
• Nat Friedman, head of product at Meta Superintelligence Labs, defended the design as intentional and scalable.
• Muse has already reached No. 1 on the U.S. App Store, making it a major consumer AI launch.
• The episode underscores how open-source projects increasingly shape commercial AI assistants.