Meta has confirmed that its new personal artificial intelligence agent, Muse, was heavily inspired by OpenClaw, an open source AI agent project. At the same time, Meta executive Nat Friedman has said that Muse was built from scratch by the company.
The clarification follows growing discussion among AI developers and early users who noticed similarities between Muse and OpenClaw. The comparisons focused not only on the way the two AI agents operate but also on elements of their internal workspace structure and configuration files.
Nat Friedman, head of product at Meta Superintelligence Labs, addressed the discussion in posts on X. He said Meta had been strongly influenced by OpenClaw and described the project as an important source of inspiration for the development of Muse. Friedman also said that the Meta team wanted to create something similar to OpenClaw while making it safer, more secure, easier to use and capable of serving a much larger user base.
According to Friedman's account, he personally used OpenClaw in January. He subsequently purchased hundreds of Mac mini computers for the Meta Superintelligence Labs team, and several members of the team began using OpenClaw and other personal AI agents.
Friedman praised OpenClaw creator Peter Steinberger and said the project's approach to building an AI agent was pioneering. He explained that Meta's objective with Muse was not simply to reproduce OpenClaw but to develop a personal AI agent that could operate at the scale required for billions of people.
The discussion became more prominent after AI app entrepreneur Ansh Nanda posted comparisons between Muse and OpenClaw. Nanda described Muse as a consumer-focused version of OpenClaw and shared a conversation in which Muse itself reportedly acknowledged similarities between its system files and those of OpenClaw. Other users subsequently examined the structure of the two systems and highlighted similarities in certain files.
One of the files that attracted attention was SOUL.md. Such a file can be used as a plain text configuration document for an AI agent, defining aspects such as personality, communication style, values, behavioural boundaries and areas of expertise. Reports about the comparison said that the file name and some of its contents appeared highly similar between the two systems.
When questioned about the similarities, Friedman did not dispute that Meta's team had adopted ideas from OpenClaw. Instead, he explained that the team believed Steinberger had made particularly effective design choices and that those ideas influenced the development of Muse. Meta's position, therefore, is that Muse was independently built while being substantially influenced by the design and approach of OpenClaw.
The development comes at a time when AI companies are increasingly moving beyond traditional chatbots toward autonomous or semi autonomous AI agents. These systems are designed not only to answer questions but also to perform tasks on behalf of users.
Meta introduced Muse publicly on September 8, 2026. The company describes it as a personal AI agent designed to help users complete tasks and manage projects. Muse can interact with connected services and, according to Meta, can perform activities such as sending emails, booking travel, creating plans and handling other digital tasks.
Meta says Muse is powered by its Muse Spark model and operates through a dedicated security environment called Muse Secure VM. The company says the system uses a separate Sentinel component to monitor and approve internet access and incorporates security controls intended to protect user information and credentials.
The company has positioned Muse as an AI assistant designed for people without specialised technical knowledge. Users can communicate with it through the Muse application and, according to Meta, through WhatsApp. The company has also said that the service is intended to work across everyday digital activities rather than being limited to question answering.
Muse has attracted significant attention since its launch. Data from mobile intelligence company Apptopia, reported by TechCrunch, indicated that Muse recorded 1.8 million iOS downloads in the United States and Canada during its first 12 days, compared with 1.3 million downloads for ChatGPT during the comparable period after its mobile launch. The comparison was limited to iOS in those markets because the two products had different initial availability.
TechCrunch also reported that Muse had reached No. 1 on the US App Store after initially launching at No. 2. Apptopia estimated that Muse had 2.8 million total installs globally during its first 12 days. These figures are third party estimates rather than numbers released directly by Meta, so they should be treated accordingly.
The rapid adoption of Muse highlights the growing competition in the personal AI agent market. Companies are developing systems that can interact with email, calendars, shopping services, travel platforms and other applications. This represents a shift from AI systems that primarily generate text toward agents that can take actions based on user instructions.
However, greater access to personal services also creates questions about privacy and security. Muse can potentially connect to services containing sensitive information, including email, payments and other personal data. Meta says it has designed Muse with a dedicated secure computing environment and that users remain in control of which services they connect.
The OpenClaw connection adds another dimension to the discussion around Muse. Meta has not said that Muse is a direct copy of OpenClaw. Instead, Friedman's comments acknowledge that OpenClaw significantly influenced the product's direction while maintaining that Meta built Muse from scratch.
For the wider AI industry, the episode illustrates how open source projects can influence commercial AI products. Open source software allows developers to examine, adapt and build upon publicly available technology, although the specific legal and technical terms governing any reuse depend on the project's licence and the nature of the material involved.
The development also demonstrates the increasing importance of AI agents as technology companies compete to build systems capable of carrying out tasks rather than simply responding to prompts. Meta's strategy with Muse is focused on bringing such capabilities to a broad consumer audience.
As Muse expands, attention is likely to remain on its performance, privacy safeguards, availability and the way Meta develops the product. The company's acknowledgment of OpenClaw's influence provides additional context about the ideas that shaped Muse, while the distinction between inspiration and direct software reuse remains important when assessing similarities between the two projects.
For Indian technology users, the development is also relevant because the global AI industry is increasingly moving toward agent based services. As these products become more widely available, consumers will have more options for using AI to manage everyday digital tasks. Their adoption will also depend on factors such as availability in India, pricing, language support, privacy controls and the services that each agent can access.
At present, the central point from Meta is clear: Muse was built from scratch, but the company acknowledges that OpenClaw had a significant influence on its product design. The disclosure follows public comparisons between the two AI systems and comes as Muse continues to gain attention in the rapidly developing personal AI agent market.