Cursor · Memtro

Cursor can talk to Memtro in two ways: as the model behind chat (so every conversation has your organisation's memory), and as an MCP server (so the agent can look things up in your systems).

Memtro as Cursor's model

Create an API key on Dashboard → API keys.
In Cursor: Settings → Models. Under OpenAI API key paste the Memtro key, turn on Override OpenAI Base URL and set it to:

https://platform.memtro.com/v1

Add a custom model named auto (or memtro, or anthropic/claude-opus-5) and select it.

Cursor's requests now go through Memtro: relevant memories are injected, facts are extracted after each reply, and auto picks a model by difficulty. Cursor's own tools keep working because Cursor executes them.

Cursor validates the key with a models request; Memtro answers GET /v1/models with the ids available to you, so the check passes.

Memtro's tools inside Cursor (MCP)

Add to .cursor/mcp.json in a project, or the global MCP settings:

{
  "mcpServers": {
    "memtro": {
      "url": "https://platform.memtro.com/api/mcp",
      "headers": { "Authorization": "Bearer sk-memtro-..." }
    }
  }
}

The agent then has memtro_search, memtro_remember, memtro_explore and your connector tools. Ask it to check the CRM, a database or a ticket as part of a coding task.

Tips

Use auto:agent as the model name when you want Cursor's chat to run a full multi-step investigation across your systems.
Memory scope follows the key: a key created inside an organisation reads and writes that organisation's shared memory (personal facts about you are still kept personal).