Perplexity Brain for Computer

The Short Version

Perplexity Brain is a self-improving memory system for Perplexity’s Computer agent. It builds a living context graph across sessions, connectors, files, projects, decisions, and sources, so the agent can start each task with relevant prior context instead of from scratch.

Unlike memory that just stores your preferences, Brain focuses on what the agent did, what worked, what failed, and what corrections were made, and updates that context at intervals, such as overnight.

What Perplexity Brain Is Best For

• Persistent agent memory: carry project context into future tasks
• Research continuity: keep a map of files, projects, and sources
• Context graph: connect related work rather than isolated chats
• Fewer cold starts: agents begin with relevant background
• Ongoing improvement: the system reviews and refines its context over time

How Perplexity Brain Differs

Most memory systems center on the user (preferences, contacts, style). Brain centers on the agent’s work and how to do it better next time. Perplexity reports gains in correctness, recall, and task cost on work that needs historical context, treat those as the vendor’s own figures.

Honest Limitations

• Research preview: features and behavior are still evolving
• Higher-tier plans: offered to Max and Enterprise Max subscribers at launch
• Memory risks apply: stored context can become outdated or sensitive; review what it keeps

Alternatives Worth Knowing

• NotebookLM, source-grounded research and document understanding
• Mem, AI note-taking with memory
• Notion, workspace and knowledge management with AI