[ ABORT TO HUD ]
SEQ. 1
SEQ. 2
Cursor Memories
Persistent Cross-Session Knowledge
Rules solve a fundamental limitation: LLMs don't retain memory between completions. Cursor Memories attack the same problem from the other direction - instead of you writing instructions, Cursor learns facts from your conversations and recalls them in future sessions. Memories replaced the earlier Notepads feature (deprecated in late 2025).
Memories vs Rules
| Aspect | Cursor Rules | Cursor Memories |
|---|---|---|
| Purpose | Prescriptive instructions | Descriptive knowledge |
| Tone | "Always do X" / "Never do Y" | "The project uses X" / "User prefers Y" |
| Authoring | You write them (.mdc / AGENTS.md) | Auto-captured from conversations |
| Review | Version-controlled files | Managed from Cursor Settings |
How to Work With Memories
- Auto-capture: when the agent learns something durable ("this repo uses pnpm, not npm"), it can store it as a memory for future sessions.
- Ask directly: tell the agent to "remember that we deploy via CLI only" - it persists the fact.
- Audit regularly: review saved memories in Cursor Settings and delete anything stale - outdated memories cause confidently wrong suggestions.
- Prefer rules for policy: anything your team must follow belongs in version-controlled rules, not personal memories.
Best Practice: Memories are personal and implicit; rules are shared and explicit. If a fact matters to anyone besides you, promote it to a
.cursor/rules/*.mdc file or AGENTS.md so it's reviewed and versioned with the code.SYNAPSE VERIFICATION
QUERY 1 // 3
What is the key difference between Cursor Memories and Cursor Rules?
They are the same thing
Rules are prescriptive instructions you write; Memories are descriptive knowledge auto-captured from conversations
Memories are only for paid plans
Rules work offline, Memories don't