Perplexity Launches “Brain” Memory System for AI Agents

Artificial intelligence company Perplexity has introduced a new memory architecture called “Brain,” designed to help AI agents learn from previous tasks and improve their performance over time. According to the announcement, the system enables agents to retain useful information, recognize patterns from earlier interactions, and apply those lessons to future assignments. The feature is currently available in research preview for Perplexity Max subscribers. The company says Brain is intended to address one of the major limitations of AI systems: the inability to build long-term contextual knowledge across multiple sessions and workflows. Traditional AI models often start each task with limited memory, while Brain aims to create a persistent layer that continuously evolves. Perplexity shared benchmark results suggesting improvements in accuracy, recall, currentness, and efficiency when Brain is enabled. The company claims the memory system can help agents provide more relevant responses, reduce repetitive work, and complete tasks with fewer prompts. The launch reflects a broader industry trend toward autonomous AI agents capable of handling complex workflows. Memory systems are increasingly viewed as a critical component for next-generation AI applications because they allow software to adapt to user preferences, project requirements, and historical interactions. Industry observers believe such developments could accelerate enterprise adoption of AI agents in research, customer service, software development, and productivity applications. However, questions remain regarding privacy, data retention, and how persistent memories will be managed securely. As competition intensifies among AI providers, innovations like Brain highlight the growing focus on creating more personalized and capable AI systems that can continuously learn from experience rather than operating solely on static training data.