Meta Says Muse Was Built From Scratch Despite OpenClaw Inspiration

Meta has acknowledged that its Muse personal AI agent was heavily inspired by OpenClaw after users flagged matching workspace files and nearly identical configuration content. Nat Friedman, head of product at Meta Superintelligence Labs, said Muse was built from scratch, framing the resemblance as product influence rather than a shared implementation.

The debate gained momentum after an AI app founder described Muse as “OpenClaw for normies” in a post that spread widely on X. Users pointed to matching names for agent-workspace files and a SOUL.md file with nearly identical content. That file is a plain-text configuration used to define an agent’s personality, communication style, values, boundaries, and expertise.

Friedman did not dispute those similarities. Asked about the shared file names and SOUL.md content, he said Meta believed OpenClaw creator Peter Steinberger had gotten those workspace elements “exactly right.” Meta’s position is therefore narrower than a denial: it acknowledges adopting OpenClaw conventions while saying it wrote Muse independently.

We built Muse from scratch, but it is definitely heavily inspired as a product by OpenClaw.

Nat Friedman, head of product at Meta Superintelligence Labs

Friedman said Meta employees used OpenClaw and wanted to create something similar that could be safe, secure, easy to use, and scale to billions of people. He said he bought hundreds of Mac minis for the Superintelligence Labs team after using OpenClaw in January, underscoring how directly the product informed Meta’s work.

Muse launched September 8 as a personal agent intended to take actions for users, including tasks such as sending email, booking travel, and making purchases. Meta says it runs in a dedicated Secure VM, or virtual machine, that holds the agent and a person’s data.

Meta also says a separate Sentinel agent reviews whether Muse’s activity can reach the internet and seeks permission when needed. Users can choose connected apps, change access, and opt out of having interactions used to train Meta AI models. Those are Meta’s stated security and privacy features, and they are central to its argument that Muse is designed for a different deployment setting than the product that inspired it.