Perplexity AI has unveiled a new artificial intelligence system called “Computer,” designed to autonomously coordinate multiple AI agents to complete complex tasks over extended periods — ranging from hours to months.
The feature is currently available to subscribers of Perplexity’s premium Max tier and is positioned as a major step toward fully automated workflow execution.
Multi-Agent, Multi-Model Architecture
According to the company, Computer allows users to define a desired outcome — such as launching a local digital marketing campaign or building an Android research application — after which the system breaks the objective into smaller tasks and assigns them to specialized AI agents.
Each agent operates using a model selected for its strengths. The reasoning core is powered by Anthropic’s Claude Opus 4.6. Google’s Gemini handles deep research, Nano Banana generates images, Veo 3.1 manages video production, xAI’s Grok supports lightweight rapid tasks, while OpenAI’s ChatGPT 5.2 is deployed for long-context recall and broad search.
The “best-model-for-the-task” strategy contrasts with offerings such as Claude Cowork, which relies solely on Anthropic’s in-house models.
Cloud-Based Execution
Perplexity said all tasks are executed in isolated cloud environments equipped with access to a real filesystem, web browser capabilities and built-in tool integrations. This structure enables AI agents to operate across files, services and applications in a coordinated manner.
The company describes Computer as a streamlined alternative to the manual orchestration advanced users often construct themselves — typically combining multiple models and protocols such as Model Context Protocol (MCP) to grant AI systems access to local tools and files.
The concept builds on earlier multi-agent orchestration experiments like OpenClaw, but Perplexity aims to broaden accessibility by packaging the automation into a unified, cloud-based platform.
The launch signals intensifying competition in the race to develop AI systems capable of sustained, autonomous task execution — a frontier many industry leaders see as the next phase of generative AI evolution.
