Meta has acknowledged that its Muse personal agent was heavily inspired by OpenClaw, while insisting the application itself was built from scratch. The distinction emerged after users compared the two systems and claimed that parts of Muse’s internal workspace closely resembled the open-source agent.
Nat Friedman, Meta Superintelligence Labs’ head of product, said his team had fallen in love with OpenClaw and wanted to build something similar that could be made secure, easy to use and capable of reaching billions of people. His statement gives the resemblance an official explanation, but it does not settle every claim about how closely the products match.
User Comparisons Focused on Agent Files
Speculation initially centered on whether Muse worked well because it was effectively OpenClaw underneath a consumer-friendly interface. An AI app co-founder amplified that theory in a viral post after a conversation with Muse described similarities in the two systems as a match rather than a coincidence.
Other users pointed to shared file names, including SOUL.md, a Markdown configuration file that defines aspects of an agent’s personality, communication style, values, boundaries and expertise. One comparison claimed that the contents of the files were almost identical.
Those observations came from users on social platforms, not from an independent technical audit described by TechCrunch. A chatbot’s answer about its own implementation is also not reliable proof of the software beneath it. The comparisons created the controversy, but they do not by themselves establish that Meta copied OpenClaw’s code.
Friedman’s response drew a narrower line. He said Muse was built from scratch but “heavily inspired as a product” by OpenClaw. When asked about identical file names and allegedly similar SOUL.md content, he did not dispute the premise and said OpenClaw creator Peter Steinberger had gotten those elements right.
Inspiration Extends Beyond a General Idea
Friedman said that after using OpenClaw in January, he bought hundreds of Mac minis for the Meta Superintelligence Labs team. Many employees began using OpenClaw and other personal agents, shaping the goal of creating a comparable experience for a much larger audience.
OpenClaw is an open-source project, which makes its architecture and conventions available for others to study. Friedman praised its creator’s agent harness as pioneering. Meta’s account therefore frames Muse as an independent implementation informed by an influential open project rather than a repackaged copy.
The near-identical file-content allegation remains more specific than ordinary product inspiration. TechCrunch reports that Meta did not offer further comment beyond Friedman’s statement. Without a code comparison or fuller technical disclosure, the available evidence supports two separate conclusions: Meta openly borrowed product ideas, and Meta says its implementation was written independently.
Keeping those claims separate matters. “Built from scratch” concerns the construction of the software, while shared structures, names and configuration language concern design choices. Both can be true, but users and open-source developers may judge originality by more than whether source code was directly reused.
Scale Changes the Stakes of an Open-Source Pattern
Muse has reached the top position in the US App Store, and third-party data cited by TechCrunch indicates that its early launch is outpacing ChatGPT under comparable conditions. Meta’s stated ambition is to take an agent pattern appreciated by technical users and make it safe and simple enough for billions of people.
That scale makes provenance more than a niche developer argument. Product conventions created in an open project can become the default experience for a mass audience when adopted by a major platform. Credit, transparency and the boundary between inspiration and imitation become part of the company’s relationship with the ecosystem that influenced it.
Meta has now clarified the strategic lineage of Muse without answering every technical question raised by the comparisons. The next useful evidence would be a more detailed account of which elements were independently designed, which conventions were deliberately adopted and how the company intends to recognize the open-source work that shaped the product.
