Memory and Brain: Perplexity That Remembers You

The Forgetting Problem

A fundamental limitation of most AI tools is that each session starts fresh. You explain your context, your preferences, and your background every time. Memory features exist to solve this: the tool learns things about you and applies that knowledge to future interactions without you having to repeat it.

Perplexity has two related but distinct memory mechanisms: the standard Memory system and Brain (which is specifically for Perplexity Computer users).

The Memory System

Perplexity's Memory system learns from your interactions over time. It picks up on your stated preferences, your expertise level, your professional context, and your recurring topics — and it applies this to future responses. If you have told Perplexity in previous threads that you are a developer who prefers concise technical answers, future responses in any thread should reflect that without you having to re-state it.

You can also pin snippets manually. If a research thread surfaces a fact, a definition, or a piece of context you want to keep accessible, pin it to memory and it will be referenced when relevant in future sessions.

Managing Your Memory

Go to Settings → Memory to see everything Perplexity has stored about you. You can:

• Review individual memory entries
• Edit entries that are inaccurate or outdated
• Delete specific entries
• Clear your entire memory if you want to start fresh

Reviewing your memory periodically is a good habit — especially if your context has changed (new job, new project, new area of interest).

Brain: Memory for the Computer Agent

Brain is a more sophisticated memory system designed specifically for Perplexity Computer — the multi-step agent we will cover in the next lesson. While standard Memory learns your preferences, Brain learns from the outcomes of agent tasks.

When you run a Computer task and correct the agent, point out a better approach, or note that a particular source was unreliable, Brain stores those corrections. The next time you run a similar task, it applies what it learned: it remembers which methods succeeded, which sources failed, and what adjustments you required.

This makes Perplexity Computer significantly better over time at your specific workflows. Brain was launched in June 2026 specifically to address the "groundhog day" problem of agentic AI — where every task run resets what the agent knows about how you work.

What Persists and What Does Not

Memory and Brain are persistent — they survive across sessions and days. Individual thread context does not persist across threads by default: each new thread starts without the content of previous threads unless you are working inside a Space with custom instructions, or a memory entry captures the relevant context. Think of Memory as your profile, and threads as individual conversations that happen within that profile.