> ## Documentation Index
> Fetch the complete documentation index at: https://docs.claude-mem.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Cursor Integration

> Persistent memory for Cursor — install with npx, no git clone

# Cursor Integration

Give Cursor a memory that lasts across chats. The installer wires Cursor hooks (small scripts Cursor runs when a session starts, a file is read, or a tool runs), starts the local worker (the program that stores notes), and uses **CMEM Pro** (the hosted memory service) unless you pick another provider.

<CardGroup cols={2}>
  <Card title="One command" icon="terminal">
    `npx claude-mem install` — do not clone the repo to install.
  </Card>

  <Card title="CMEM Pro default" icon="cloud">
    Memory notes are written off-plan through the hosted service unless you pick Gemini, OpenRouter, or your Anthropic plan.
  </Card>

  <Card title="Cursor hooks" icon="bolt">
    Cursor has hooks. Grok Bot does not — that host is a [different guide](/grok-bot).
  </Card>

  <Card title="Memory viewer" icon="eye">
    Open the worker URL printed at install to browse saved notes.
  </Card>
</CardGroup>

## Install

```bash theme={null}
npx claude-mem install
```

When the installer lists your apps, pick **Cursor**. The three stages are: install runtime, sign in (device code, no email in the CLI), then pick a memory provider. **CMEM Pro is pre-selected.**

Other provider choices: your Gemini key, your OpenRouter key, or your Anthropic plan (`--provider claude`, which skips cmem.ai). Details: [Installation](/installation).

<Warning>
  `npm install -g claude-mem` installs the library only. It does **not** register Cursor hooks or start the worker. Always use `npx claude-mem install`.
</Warning>

Do not clone this repository to install. Clone is for people changing the code.

## How it works

1. **Hooks** record what the agent does in Cursor.
2. **Worker** stores those notes on your machine (default `127.0.0.1`, port in `~/.claude-mem/.worker.port`).
3. **Next chat** gets relevant past notes.
4. **Viewer** at the worker URL shows the knowledge base.

Never restart a healthy worker. The note queue sits in memory; a restart drops it.

## Gemini or OpenRouter instead of CMEM Pro

* [Gemini](/cursor/gemini-setup) — your Google key, memory off-plan
* [OpenRouter](/cursor/openrouter-setup) — your OpenRouter key. **Never** send a personal `sk-or-` key to `https://cmem.ai/api/inference`

## Verify it worked

1. Restart Cursor after install so hooks load.
2. Health: `curl -s "http://127.0.0.1:${CLAUDE_MEM_WORKER_PORT}/api/health"` (or the port in `~/.claude-mem/.worker.port`).
3. Do a small unit of work, then open the worker URL. New notes should appear.

## Troubleshooting

### Nothing is being stored

* Confirm the worker is up. Do **not** restart it if it is already healthy.
* Confirm you used `npx claude-mem install`, not a global `npm install`.
* Restart Cursor once after install.
* Logs: `~/.claude-mem/logs/worker-YYYY-MM-DD.log`

### Hooks not firing

Restart Cursor. Hooks live under `.cursor/` in the project or your user Cursor config after install.

## Next steps

* [Installation](/installation) — installer stages and CMEM Pro
* [CMEM Pro (manual / headless)](/cmem-pro-headless) — settings the installer writes
* [Gemini](/cursor/gemini-setup) / [OpenRouter](/cursor/openrouter-setup)
* [Search Tools](/usage/search-tools) — query project history