Perplexity has introduced a new agentic AI system called Perplexity Computer, positioning it as a kind of digital coworker capable of handling complex, multi-step projects.
The tool operates by deploying multiple subordinate AI agents that collaborate to complete tasks. Instead of responding to a single prompt with a single output, Perplexity Computer can break projects into smaller pieces, assign them to specialized sub-agents, and then assemble the final result.
For example, it can build dashboards, generate presentations, develop simple apps, or compile research-heavy reports. According to Perplexity, the system leverages several leading AI models simultaneously—using Claude Opus for deeper reasoning, Gemini for research tasks, and other models for image generation, video output, and faster subtasks.
Unlike open-source agentic systems such as OpenClaw, Perplexity Computer runs entirely in the cloud within a managed environment. This design reduces the risk of the AI interacting directly with a user’s local files or operating system. However, the tradeoff is reduced flexibility compared to locally run agents that can directly manipulate desktop applications and system-level workflows.
Access to Perplexity Computer is currently limited to subscribers of the Perplexity Max plan, which costs $200 per month. The premium pricing reflects the system’s multi-model architecture and higher compute demands.
As agentic AI tools continue to evolve, Perplexity’s latest launch highlights a growing split in the market between cloud-controlled “AI coworkers” and locally operated, user-managed AI agents.
