Perplexity Launches Autonomous AI Agent
AI firm Perplexity has launched "Perplexity Computer," an AI system that can create and execute entire workflows on its own. It's a major step in "agentic" AI, where AI agents can orchestrate complex business processes without direct human supervision.
The system orchestrates 19 specialized AI models, including from Google and Anthropic, to autonomously execute complex work. Perplexity Computer's internal framework decomposes a user's goal into subtasks, then routes each piece to the most effective model for that specific job, such as coding, research, or design. This new agent operates entirely in the cloud within an isolated computing environment, a design choice meant to enhance safety. Unlike AI agents that run locally on a user's machine with deep file access, this approach prevents the agent from taking unintended actions on a personal device. Access to Perplexity Computer launched on February 25, 2026, exclusively for subscribers of its $200/month "Max" tier. The service uses a credit-based system for tasks, with plans to expand access to its $20/month "Pro" and Enterprise customers in the near future. The launch represents a significant pivot for Perplexity AI, moving from a "knowledge engine" focused on answering questions to an "action engine" designed to deliver completed projects. The company, founded in August 2022, has seen its valuation soar to $20 billion as of September 2025, backed by investors like Jeff Bezos and Nvidia. Perplexity was founded by CEO Aravind Srinivas, a former researcher at OpenAI and Google Brain, along with Denis Yarats from Meta AI, Johnny Ho from Quora, and Databricks co-founder Andy Konwinski. By late 2025, the company was approaching a $200 million annualized revenue run rate from its subscription services.
