Small publishers lost 60% of search referrals. Google beat a publishers antitrust suit. For brands, the game is becoming a source inside the answer layer.
Perplexity's Personal Computer turns a Mac mini into a persistent, always-on AI agent that works across your local files and apps around the clock — a new category of ambient computing.
Perplexity just unveiled Personal Computer, and it feels like a new category of computing — an always-on AI that sits next to your work, not just in a browser window.
Personal Computer is a software-driven AI agent that runs 24/7 on a continuously powered Mac mini, merging your local apps, files, and sessions with Perplexity's cloud-based intelligence and agent platform. It is not a new gadget. It is software that turns a Mac mini into a persistent, remote-controllable AI assistant designed to act on your behalf across workflows.
What Personal Computer Actually Does
The core proposition is persistent local access combined with cloud AI reasoning. Traditional AI assistants — ChatGPT, Claude, Gemini — run in a browser tab. They have no access to your local file system, no awareness of your running applications, and no ability to interact with your desktop environment. When you close the tab, they are gone.
Personal Computer solves that by running continuously on a dedicated Mac mini. It can read, search, and interact with your files and applications while you are away, combining that local context with Perplexity's cloud-based AI for complex reasoning. According to 9to5Mac's coverage, the system functions as "an always on, local merge with Perplexity Computer that works for you 24/7."
The product integrates with the tools knowledge workers already use: Gmail, Slack, GitHub, Notion, and Salesforce. It can monitor triggers, execute proactive tasks, and carry work forward around the clock without requiring the user to be present. You describe objectives rather than step-by-step instructions, and the system handles execution.
Crucially, it is controllable from any device, anywhere. Your Mac mini stays at home or in the office running continuously, and you reach it remotely — a persistent digital proxy that maintains context across sessions.
Safety Controls Are Central to the Design
An always-on AI agent with access to your local files, email, and code repositories raises obvious security questions. Perplexity has made safety controls a prominent part of the product design rather than an afterthought.
Three mechanisms are built in:
User-approval prompts. Sensitive actions require explicit approval before the agent can execute them. The agent cannot silently send an email, push code, or delete files without your sign-off.
Comprehensive action logging. Every action the agent takes is logged and auditable. You can review exactly what it did, when, and why.
A kill switch. Immediate stoppage. If the agent is doing something you did not intend, you can halt it instantly.
Sessions run in isolated Firecracker microVMs, the same lightweight virtualization technology that AWS uses for Lambda and Fargate. This reduces risk when the agent handles untrusted content or sensitive data by isolating each session from the host environment.
As Macworld noted, the system includes "clear safeguards" with every action logged and approval gates for anything consequential. Whether these controls prove sufficient at scale remains to be seen, but the architecture takes the right starting position.
Hardware and Pricing
The hardware choice is deliberate. Apple's Mac mini — likely the M4 configuration — is compact, energy-efficient, and designed to run continuously. It is already a popular choice for home servers and always-on workloads, which makes it a natural platform for a persistent AI agent.
Personal Computer requires a Perplexity Max subscription, the company's highest tier, priced at $200 per month. The service is Mac-only at launch. Perplexity has not disclosed whether it will supply pre-configured Mac mini units or expect users to provide their own hardware, though AppleInsider reports the expectation is that users dedicate a spare Mac mini to the role.
The Waitlist Signal
Access today is limited. Perplexity opened a waitlist on March 11, 2026, rather than launching broadly. The company says it will "provide support and resources for the initial cohort," signaling a cautious, staged rollout rather than a rush to scale.
This is the right approach for a product with this much surface area. An always-on agent with local file access and integration into email, messaging, and code repositories needs to earn trust incrementally. A broad launch with unresolved edge cases would be a reputational risk that Perplexity cannot afford as it competes with larger, better-resourced AI companies.
Why This Matters for AI Search and Discovery
Personal Computer is not a search product in the traditional sense. But it extends the logic that made Perplexity relevant in the first place: AI should not just answer questions — it should do work on your behalf, grounded in real data.
The always-on agent model creates a new relationship between users and AI. Instead of opening a chat window when you have a question, the AI is continuously available, aware of your local context, and able to act proactively. It can monitor your inbox for relevant messages, track changes in codebases, surface research while you sleep, and prepare summaries before you sit down in the morning.
This is closer to the original promise of AI assistants than anything currently in market. Whether Perplexity can deliver on that promise — at $200 a month, through a waitlist, on a single hardware platform — is the open question. But the category itself feels inevitable. If Perplexity does not build the always-on AI agent that works, someone else will.
James Calder is the editor of The Search Signal, covering AI-powered search, generative engine optimization, and the future of brand discovery.
Small publishers lost 60% of search referrals. Google beat a publishers antitrust suit. For brands, the game is becoming a source inside the answer layer.
Brave Search now offers two AI-native API endpoints that deliver pre-extracted, relevance-scored web content directly to LLMs — signaling a fundamental shift in how search APIs serve machine consumers.
