CognitiveXCOGNITIVEX · INTEGRATIONS
Persistent memory for Cursor
Cursor removed its built-in Memories in v2.1. CognitiveX gives Cursor durable, cross-session memory over MCP, so your editor remembers decisions, conventions, and project context across every chat, not just the one you're in.
WHAT HAPPENED TO CURSOR MEMORIES
Cursor forgets between chats. The LCM doesn't.
Add one MCP server and your editor gets a memory that survives the session.
Cursor shipped an experimental Memories feature and then pulled it as a built-in in the 2.1 release, leaving most workflows back where they started: each new chat begins from a blank slate. Cursor is excellent at reasoning over the code in front of it, but by design it doesn't carry why you made a decision last Tuesday, which library you standardized on, or the naming convention you agreed three repos ago.
CognitiveX fills exactly that gap. It's the Large Cognition Model, a cognitive layer that sits beside any LLM and gives it a persistent, evolving memory. You connect it to Cursor through the Model Context Protocol (MCP), and from then on Cursor can remember and recall across every chat, every project, and every restart. The memory is the model, and it lives outside the editor, so closing the window doesn't erase it.
TWO-MINUTE SETUP
Wire CognitiveX into Cursor over MCP
Cursor speaks MCP natively, so there's nothing to fork or patch. You add the CognitiveX server, point it at your API key, and reload.
Get a CognitiveX API key
Sign in and create a key. The free Amnesiac tier is enough to try it (100 recall credits/month, 1,000 memories; storing memories is always free). Keys and quotas live in your dashboard. Full reference is in the docs.
Add the MCP server to Cursor
Open
Cursor Settings → MCP → Add new server, or drop a.cursor/mcp.jsonin your project (or~/.cursor/mcp.jsonfor all projects):{ "mcpServers": { "cognitivx": { "url": "https://api.cognitivx.io/mcp", "headers": { "Authorization": "Bearer YOUR_COGNITIVX_KEY" } } }}
Reload and confirm the tools
Reload Cursor. In the MCP panel you should see the CognitiveX tools light up:
remember,recall,reflect, and the rest. A green dot means the handshake succeeded.Teach it once, recall it forever
In any chat, tell Cursor something worth keeping: “remember we chose Drizzle over Prisma for the edge runtime”. Open a brand-new chat tomorrow and ask “what's our ORM decision and why?” Cursor recalls it through CognitiveX. The context you build is no longer trapped in one conversation.
Prefer to wire it into your own tooling instead of (or alongside) the editor? The same memory is reachable from the cogx SDK and HTTP API. See the developer docs.
WHAT YOU GAIN
Cursor, with a memory that compounds
| Capability | Cursor alone | Cursor + CognitiveX |
|---|---|---|
| Remembers within a single chat | ||
| Built-in Memories feature | ||
| Persists across new chats & restarts | ||
| Shared across projects & repos | ||
| Same memory in Claude Code / Codex / ChatGPT | ||
| Consolidates episodes into durable facts | ||
| Pattern detection & overnight consolidation |
The deeper difference: most "memory" add-ons just store text and search it back. The CognitiveX LCM consolidates. Episodic events (what you did) get promoted into semantic facts (how this codebase works); patterns surface on their own; an overnight dream-consolidation pass reorganizes the substrate so recall stays sharp instead of drowning in stale notes. Cursor gets a memory that gets better the more you use it, across all four tiers: semantic, episodic, procedural, and foundational.
NOT JUST CURSOR
The same memory follows you everywhere.
Because CognitiveX connects over MCP (and through the cogx SDK and HTTP API), the memory you build in Cursor isn't locked to Cursor. Switch to Claude Code for a refactor, ask Codex or ChatGPT a question, or hit the API from a CI job: they all read and write the same living memory. Decide something in your editor in the morning and your agent already knows it in the afternoon.
- cross-session: survives new chats, reloads, and machine restarts
- cross-tool: one memory shared across every MCP client you use
- cross-repo: conventions and decisions aren't trapped in one project folder
- storing is free: you only spend recall credits when you read memory back
See the full picture on the integrations overview, read how the loop works on the LCM page, or weigh it against alternatives on the comparison page.
QUESTIONS
Persistent memory for Cursor: FAQ
Did Cursor really remove Memories?
Cursor shipped Memories as an experimental, beta feature and then removed it as a built-in in the 2.1 release. Many teams that relied on it now have no persistent memory inside the editor. CognitiveX restores that and goes further, because the memory lives outside Cursor and is shared across your other tools.
Does this work with the latest Cursor?
Yes. CognitiveX connects through Cursor's native MCP support, so it works regardless of whether a given Cursor version ships its own memory feature. If Cursor reintroduces Memories later, the LCM still adds cross-tool, cross-repo persistence and consolidation on top.
Where is my memory stored, and is it private?
Your memories live in your CognitiveX account behind your own API key, not inside the editor session and not in a shared model. You control what's stored and can recall, update, or forget individual memories. See the privacy page for specifics.
What does it cost?
Storing memories is always free. You spend recall credits only when you read memory back, priced by depth: foundational recalls are 1 credit, standard 3, deep 10. The free tier includes 100 recall credits a month, which is plenty to try it inside Cursor. Paid tiers and pay-as-you-go scale from there.
Is this just a notes file the model reads?
No. CognitiveX is a structured cognitive layer: a four-tier memory with pattern detection, salience scoring, and overnight consolidation that promotes events into durable facts. We treat the LLM as infrastructure; the intelligence is in the memory architecture, which is why recall stays useful as it grows instead of becoming a wall of stale text.