---
title: Users FAQ
description: "Everyday questions for people using Matrix — clarifying questions, corrections, memory, long conversations, and reviewing what the agent will do."
---

> **For AI agents:** the complete documentation index is at [llms.txt](/llms.txt). Append `.md` to any page URL for its markdown version.

<AccordionGroup>
  <Accordion title="Why does Matrix sometimes ask a clarifying question?">
    The active skill knows what it needs. When a required detail is missing it asks one focused question rather than guessing — this is by design, to keep your intent intact.
  </Accordion>
  <Accordion title="Can I correct the agent without starting over?">
    Yes. Because the plan is typed, you can adjust a constraint, change a target, or raise a budget structurally instead of rewriting your whole request.
  </Accordion>
  <Accordion title="Does Matrix remember things between conversations?">
    Yes — cortex stores your goals, preferences, facts, and past turns per actor, and surfaces the relevant pieces automatically so you don't repeat yourself.
  </Accordion>
  <Accordion title="What happens in a very long conversation?">
    Matrix consolidates older history into a running summary while preserving every high-entropy identifier (file paths, addresses, IDs) verbatim. You'll see a brief notice when it does this.
  </Accordion>
  <Accordion title="Can I see what the agent will do before it acts?">
    For monetary or irreversible work, the rigorous rail produces a reviewable typed plan first. You can inspect and correct it before anything runs.
  </Accordion>
</AccordionGroup>