Perplexity Turns Mac mini Into Local AI Agent

Perplexity is shipping Personal Computer, a Mac-specific expansion of its Perplexity Computer agent that runs persistently on a dedicated Mac mini and integrates with local files, native apps, and the browser. Activated by pressing both Command keys, the agent accepts text and voice commands, can run 24/7, and orchestrates work across multiple frontier models. Personal Computer exposes actions for auditability and reversal, sandboxes file creation, and supports remote initiation from an iPhone using 2FA. The feature rolls out to Perplexity Max subscribers at $200/month and waitlist members; it is not included in the $20/month Pro tier. For practitioners, this is a pragmatic step toward hybrid, locally-attested AI agents that combine on-device access with cloud model routing and managed orchestration.
What happened
Perplexity released Personal Computer, a Mac-specific extension of Perplexity Computer that turns a dedicated Mac mini into an always-on AI agent with local file and app access, multi-model orchestration, and auditable actions. The feature is available to Perplexity Max subscribers (priced at $200/month) and waitlist members; the company is not enabling it for the $20/month Pro tier yet. Activation is immediate from the Mac app by pressing both Command keys and supports text and voice input.
Technical details
Personal Computer brings Perplexity's managed orchestration to the local machine while retaining connections to Perplexity's secure servers. It can search, read, and write files in designated folders and interact with native apps including iMessage, Mail, and Calendar. Key runtime and interface characteristics include:
- •Persistent runtime on a dedicated Mac mini, enabling background, 24/7 operation for long-lived tasks and monitoring.
- •User-facing controls: visible action logs, auditable and reversible steps, a kill switch, and sandboxed file creation to limit blast radius.
- •Cross-device initiation and security: tasks can be started from an iPhone using two-factor authentication to authorize local operations.
- •Multi-model orchestration: the product routes subtasks across an ensemble of frontier models; Perplexity previously described multi-model routing in Perplexity Computer and third-party reviews indicate the platform assigns different models for discrete subjobs (examples referenced include Opus, Gemini, Grok, and GPT-5.2).
Context and significance
Perplexity is taking a hybrid approach: keep centralized model selection and skill playbooks while granting the agent direct, authenticated access to local state. That design contrasts with purely cloud-only assistants and with fully local, single-model tools like competing local agents. For practitioners this matters because it operationalizes several trends simultaneously: on-device persistence, multi-model task routing, and auditable automation. The model routing strategy reduces need for user configuration while enabling specialization across submodels, which can boost efficiency and accuracy on compound tasks such as multi-document synthesis, automation of local file workflows, and cross-referencing local files with web sources.
Practical trade-offs
The offering is powerful for Mac-centric power users and teams but has limits that matter to deployers. The feature is gated behind a high-tier subscription, raising cost and scaling questions for enterprises. Local availability is currently macOS-focused (macOS 14 Sonoma or later) and optimized for a dedicated Mac mini. Security controls (sandboxing, action logs, kill switch, and 2FA) mitigate but do not eliminate risks from granting an agent write and app-level access; organizations will need to validate threat models and compliance requirements before broad rollout.
What to watch
Track expansion beyond macOS, developer APIs for custom connectors and enterprise management, and Perplexity's roadmap for offline or fully local inference. Also watch how competitors evolve hybrid on-device orchestration and whether enterprises demand policy controls, RBAC, and audit integrations for governance.
Scoring Rationale
This is a notable product release that advances hybrid, on-device agent workflows and multi-model orchestration, which matters to practitioners building automation and local-AI deployments. It is not a paradigm-shifting frontier model release, and adoption is limited by macOS focus and high subscription pricing.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problemsStep-by-step roadmaps from zero to job-ready — curated courses, salary data, and the exact learning order that gets you hired.

