Perplexity launches its Personal Computer agent on Mac

Perplexity opens its “Personal Computer” agent to all Mac users

Perplexity is taking a new step in the race for AI agents capable of acting directly on the user’s machine. As reported by TechCrunch, the company is now making its Personal Computer agent accessible to everyone on Mac, following a more limited rollout phase. Behind this announcement, this is not merely a new feature for Perplexity subscribers: it is also another sign of the gradual mainstreaming of so-called “computer use” tools, agents capable of manipulating applications, carrying out tasks, and interacting with a real software environment.

The issue extends far beyond a product update. For several months, the major names in generative AI have been seeking to move beyond the simple conversational chatbot to offer assistants that no longer just answer, but act. In this landscape, Perplexity, initially identified as an AI-powered answer engine, is attempting to position itself in an area where OpenAI, Google, and several specialized startups are also moving quickly.

From answer engine to executing agent

The broad launch of Personal Computer on Mac is part of a wider transformation of the market. During the first wave of generative AI, value was concentrated in producing text, images, or code. Since late 2024 and early 2025, the industry has sought to move up a level: making the assistant an operator capable of carrying out concrete actions in existing interfaces.

In practical terms, these agents can navigate windows, click buttons, fill out forms, search for information across several applications, or chain together simple operations without constant user intervention. The stakes are considerable: an assistant that understands a request but leaves the user to do everything else does not have the same economic impact as an agent capable of handling part of everyday digital work.

Perplexity is not alone in this segment. OpenAI has strongly promoted its vision of action-oriented agents, notably with Operator and the gradual integration of more execution-focused capabilities around ChatGPT. Google, for its part, is pushing similar uses through its Gemini ecosystem and its work on task automation. Alongside the giants, several startups are specializing in computer automation, with promises ranging from personal productivity to the execution of business workflows.

What Perplexity is specifically announcing for Mac

According to TechCrunch, Perplexity is therefore making Personal Computer available to all Mac users, after a period of more limited access. The tool is designed to allow the assistant to carry out actions directly on the computer, rather than merely provide a text response or web summary. This broader availability on macOS is important because it moves the product from experimental status, or access reserved for certain profiles, to wider availability—and thus to testing at real scale.

The appeal of a Mac launch is not insignificant. In tech, design, product, media, or development roles, the Apple environment remains highly prevalent. For Perplexity, reaching Mac users means targeting an audience especially inclined to quickly try new automation tools. It is also favorable ground for demonstrating concrete use cases: web research, document organization, manipulation of work applications, content preparation, or assistance with repetitive tasks.

The very wording of Personal Computer is revealing. Where many assistants still present themselves as copilots, Perplexity adopts terminology that suggests a more direct relationship with the machine. The computer is no longer merely the AI’s medium: it becomes the agent’s field of action.

By making Personal Computer available to everyone on Mac, Perplexity aligns its product with the sector’s central trend: moving from AI that advises to AI that executes.

Why this rollout matters against ChatGPT, Operator, and Google

The announcement may appear modest at first glance, but it comes at a time when competition is being reshaped. Until now, Perplexity was primarily compared with ChatGPT, Gemini, or Claude in the area of assisted research, information synthesis, and conversational responses. With Personal Computer, the comparison shifts to another category: that of operational assistants.

Against ChatGPT, the stakes are strategic. OpenAI benefits from an advantage in brand recognition, a massive user base, and an already established ecosystem. But the battle is no longer being fought solely over the model’s raw quality. It is also about the ability to become the interface that saves time every day. If Perplexity can turn its expertise in research and web browsing into actions carried out on a workstation, it can defend a credible position in professional uses.

Operator, at OpenAI, has specifically helped popularize the idea of an agent capable of manipulating a digital environment as a human would. The rollout of Personal Computer on Mac shows that this logic is no longer reserved for a few premium demonstrations or limited-access programs. It is becoming a standardized product direction that several players are seeking to industrialize.

Google, for its part, has a structural advantage: the depth of its software integration, spanning browser, office tools, mobile, and cloud. But this strength can also become a constraint if the agentic experience remains too centered on its own ecosystem. Perplexity may try to differentiate itself through an approach that is more cross-functional, more use-oriented than platform-oriented.

For startups specializing in agents, pressure is rising. When a player already known to the broader tech public, such as Perplexity, widely opens up a “computer use” feature, it helps make this category more visible, but also more competitive. The effect is twofold: demand increases, while tolerance for approximation declines.

A turning point for practical uses, including in Europe

For a French and European audience, this development is particularly significant. The local generative AI market remains highly attentive to use cases that can actually be deployed in companies: support, customer relations, document research, office productivity, and automation of administrative tasks. Agents capable of using a computer can meet this expectation more directly than traditional chatbots.

In many organizations, the obstacle is no longer merely producing text or ideas, but integrating with existing tools without rebuilding the entire infrastructure. An agent that can interact with a Mac interface, open an application, retrieve information, launch a series of actions, and return a result can become an immediately useful building block. This is particularly true for SMEs, product teams, consulting firms, media outlets, or certain digital marketing roles.

In France, where the adoption of AI in business is progressing but often remains governed by strong security and compliance requirements, this type of agent also raises very concrete questions:

• what permissions does the user grant the agent on their machine?
• what data is visible, processed, or potentially transmitted to the provider?
• what limits are imposed on actions that can be carried out automatically?
• what traceability is available to audit what the agent has done?

These issues are all the more sensitive in Europe because the regulatory framework, between the GDPR and the growing implementation of the AI Act, is pushing vendors to clarify how their systems actually work. As agents move from suggestion to action, governance requirements increase mechanically.

The mainstreaming of “computer use” is only beginning

Perplexity’s opening of Personal Computer on Mac illustrates a shift in phase. The market is gradually entering a period in which agents will no longer be assessed solely on their language capabilities, but on their operational reliability. Knowing how to click, navigate, and execute is not enough: it must be done robustly, quickly, understandably, and safely.

The next battle will likely focus on several criteria. First, execution quality: an agent must complete its tasks with a low error rate, even in changing interfaces. Next, trust: the user must be able to take back control, validate sensitive steps, and understand the decisions made. Finally, integration: the agents that fit best into existing software and workflows will have a decisive advantage.

For Perplexity, this opening on Mac is therefore less an end point than a real-world test. If the company succeeds in converting its base of curious users into regular use, it will be able to reposition itself as a credible player in practical agentic AI, beyond conversational search. If the experience remains confined to impressive but unreliable demonstrations, it will join the long list of promises still awaiting maturity.

In the short term, this announcement above all reinforces one certainty: the future of AI assistants will increasingly depend on their ability to act in the real software world. And in this area, the broad availability of an agent such as Personal Computer on Mac helps normalize an idea that, not long ago, belonged to the realm of prototypes: the personal computer is gradually becoming a space for delegating to artificial intelligence. For advanced users as well as software vendors, the question is no longer really whether these agents will become widespread, but how quickly they will become reliable enough to establish themselves permanently in everyday uses.