Skip to content

Assistant Memory

The Cowriter assistant remembers durable facts about you — your field, your languages, your style preferences — so you don’t repeat yourself. On Pro and Max it can also recall your entire past-conversation history when relevant.

Assistant Memory has two layers:

  1. Memories (all plans, on by default) — durable facts the assistant learns about you, like “writes in an academic style”, “first language Russian, writing in English”, or “prefers British spelling”. These are auto-extracted from your conversations, stored, and quietly injected into the assistant’s context so it adapts to you across sessions. You can see and edit every memory in Settings → Memory.

  2. Full recall (Pro and Max only) — the assistant can semantically search your entire past conversation history and pull in relevant snippets when they’d help, Cursor-style cross-session recall.

  • Memories are automatic — there’s nothing to turn on. View, edit, pin, or delete them in Settings → Memory.
  • Full recall activates automatically inside Cowriter for Pro/Max users when past context is relevant (via the assistant’s tool).
PlanMemories (Layer 1)Full recall (Layer 2)
FreeYesNo
SparkYesNo
ProYesYes
MaxYesYes

Memories are available to everyone. Full cross-session recall is gated to Pro/Max — on Free/Spark, asking the assistant to recall older conversations returns an upgrade prompt.

  • Auto-extracted shortly after a session (on idle).
  • Near-duplicate memories are merged.
  • There’s a cap on how many are kept; the least-used, lowest-confidence ones are pruned first — but pinned memories are never auto-removed.
  • You can pin, edit, or delete any memory yourself.

Memory is private to your account. In Settings → Memory you can toggle “Don’t remember me” — this stops new extraction and recall and wipes your stored memories. Memory data is included in your data export and erasure requests.

  • Full recall is Pro/Max only. Free/Spark get durable Memories but not cross-session conversation recall.
  • Opting out clears existing memories — it’s not a pause, it’s a wipe.
  • The assistant decides which memories are relevant; you steer it by pinning the ones that matter and deleting ones that don’t.
  • “How does it know things about me without me telling it?” — It auto-extracts durable facts from your conversations (your field, languages, style) and stores them as editable memories.
  • “Can I see and delete what it remembers?” — Yes, everything is listed in Settings → Memory; edit, pin, or delete any of it.
  • “What’s the difference between Memories and full recall?” — Memories are a short list of durable facts injected every time. Full recall (Pro/Max) searches your entire conversation history on demand for relevant moments.
  • “How do I turn memory off?” — Settings → Memory → “Don’t remember me”. This also wipes what’s stored.
  • “Is my memory shared with anyone?” — No. It’s scoped to your account and covered by export/erasure.

Was this article helpful?