What Perplexity Computer Actually Is
Perplexity Computer is the autonomous agent layer that Perplexity has built on top of its search foundation. Unlike a standard Perplexity query which returns a one-shot answer, Computer can take multi-step tasks, use tools, read and write files, run browser actions, and execute automations that span multiple applications over time. It is the difference between asking a question and delegating a task.
In recent months, Perplexity has expanded Computer significantly: it is now available across Microsoft 365 apps including Word, Excel, PowerPoint, Outlook, and Teams. It can publish websites, run subagents in parallel, execute automations triggered by schedules or events, and as of October 2026, generate videos using MiniMax H3 and ByteDance Seedance 2.5 for Pro and Max subscribers.
Computer vs Standard Perplexity: When to Use Which
Use standard Perplexity for quick factual queries — anything that needs a single answer from web sources with citations. Use Computer when the task involves multiple steps, requires actions in external tools, or needs the AI to maintain context across a longer workflow.
The clearest signal you need Computer rather than standard search: if you find yourself copying answers between apps, running the same sequence of steps repeatedly, or spending time on research that could be systematized, Computer is designed to eliminate that category of work.
How Automations Work in Perplexity Computer
Introduced in September, Perplexity Automations let you define workflows that run on a schedule or are triggered by specific events. Automations use Computer memory, Skills, and connected apps including Slack, Gmail, Outlook, and Linear. You set the trigger — time-based or event-based — define the steps the automation should take, and Computer executes them without further input.
The practical applications for knowledge workers are immediate: automate a daily briefing that pulls from your most-used research sources, set up a weekly summary of mentions of your company or competitors, or create a workflow that monitors specific data sources and alerts you when conditions are met.
The Model Council: Choosing the Right Model Mid-Task
Computer now includes a Model Council feature that can switch models mid-task based on what each model does best. Rather than committing to a single model for the entire task, you can let Computer select the most appropriate model for each step — using a fast model for simple queries, a reasoning model for complex analysis, and a specialized model for domain-specific tasks.
For research workflows, this means Computer can use a fast model for initial source gathering and a reasoning model for synthesis without manual intervention.
Shared Workspaces and Projects
Computer now supports shared workspaces where team members can collaborate on research projects. Shared files and the Brain — Perplexity persistent memory system — are accessible across the workspace, making it possible to build shared knowledge bases that compound over time rather than existing only in individual conversations.
For teams, this changes how research is stored and reused. A workspace for market research, competitive analysis, or product development becomes a persistent resource that accumulates institutional knowledge rather than starting fresh in each session.
Enterprise Controls and API Access
For enterprise users, role-based access controls let administrators define who can access which workspaces and what data Computer can read and write. API credentials are now stored in a credential vault rather than embedded in code. Deep Research capabilities include a command panel, forking, inline actions, and analytics APIs for teams building automated research pipelines at scale.