Perplexity is making its Computer AI agent more private by allowing it to run entirely on supported Apple silicon Macs. Announced on October 20, 2026, the new fully local mode runs the AI model, task planning, reasoning, and execution on the Mac, allowing users to work with sensitive files and native apps without sending that information to the cloud.

The update builds on Perplexity’s efforts to turn Computer into an AI agent that can do more than answer questions. With the appropriate permissions, it can work across apps such as Notes, Messages, Mail, and Numbers, bringing together information from different sources to complete multistep tasks.

Unlike a conventional chatbot that primarily generates responses, Computer can use local files and supported applications to help carry out actual work. For example, it can review recent messages and emails, identify outstanding requests and deadlines, and create an action list in Notes. It can also use information from spreadsheets and project documents to prepare status updates and other work materials.

Perplexity Computer now offers fully local and hybrid modes

The new option expands the execution choices available in the Perplexity Mac app. Users can choose fully local processing or use Hybrid Compute, which combines cloud-based AI capabilities with local processing of private information.

In fully local mode, the model and the agent coordinating the task both run on the Mac. Perplexity says this keeps planning, reasoning, and file processing on the device, giving users greater control over sensitive information and when they use cloud resources.

The feature launches with two local model options: Qwen 3.6 and PPLX Pro, a version of Qwen further trained by Perplexity. Rather than requiring users to configure a separate local AI environment, the models can be downloaded and set up through the Perplexity Mac app.

Hybrid Compute remains useful for tasks that need more than locally stored information. It combines cloud-based research and reasoning with on-device processing of private files. For instance, a user could analyze a locally stored financial portfolio using downloaded market data, then authorize cloud assistance to research current market developments.

Users running Computer locally can also authorize cloud access when a task requires up-to-date information, browser use, or connected apps. Perplexity says its local orchestrator can call on more than 15 frontier AI models when a task needs more advanced reasoning.

This flexibility matters because fully local execution and cloud-assisted AI serve different purposes. Local processing can help keep confidential information on a personal computer, while cloud models remain useful for research and tasks that demand broader capabilities.

What can Computer do with native Mac apps?

Perplexity’s goal is to make Computer an assistant capable of coordinating work across applications rather than handling each task in isolation.

For example, it could check a project’s to-do list in Notes, find a colleague’s budget request in Messages, compare planned and actual spending in a spreadsheet, and prepare a status update for review. It can also help draft client proposals from confidential discovery notes, project briefs, and internal pricing information.

Other potential uses include reviewing correspondence to identify deadlines, collecting relevant documents, drafting follow-up messages, and organizing meeting-related tasks.

Perplexity says code and tool execution take place in isolated sandbox environments with controlled access to files and connected apps. The company says the same security protocols are used in its cloud-based Computer experience.

However, running AI locally does not automatically eliminate every security risk. Users still need to consider the permissions they grant an agent and review its actions, particularly when working with sensitive documents or communications. The privacy benefit is that fully local processing reduces the need to transmit the underlying information to remote AI servers.

Availability and Mac requirements

Perplexity Computer’s fully local mode is available through the Perplexity Mac app for Pro, Max, and Enterprise subscribers with supported Apple silicon Macs.

The company recommends 36GB of unified memory or more for running Computer locally. That recommendation makes memory capacity an important consideration for anyone planning to use an AI agent for demanding workflows on a Mac.

To get started, users need to install or update the Perplexity Mac app, follow the in-app instructions to download and configure a local model, and select the fully local option when starting a task.

The launch is another step toward making AI agents capable of handling sensitive work directly on personal computers. Instead of relying entirely on remote AI infrastructure, Mac users can now choose to keep an entire workflow on their device or combine local processing with cloud-based capabilities when a task calls for them.