// integration

Add memory to Cursor

Cursor speaks MCP, so giving its agent persistent memory is a few lines in an mcp.json file. Add the server, reload, and Cursor can store and recall named memory across sessions and projects.

In one line: Add AgentRAM's MCP server to Cursor and its agent can store and recall persistent memory as native tools.

What you get

Persistent memory that survives across Cursor sessions and projects: named values you store once and recall later, shared memory for multi-agent work, temporal memory for facts that change over time, and no vector database to run. It is fifteen tools in total, and they show up as tools Cursor's agent calls directly. For the concepts underneath, see the MCP integration overview.

Prerequisites

Add the server

Cursor reads MCP servers from an mcp.json file. Use a global file to make AgentRAM available in every project, or a project file to scope it to one repo:

Add this block to whichever file you chose:

{
  "mcpServers": {
    "agentram": {
      "command": "npx",
      "args": ["-y", "agentram-mcp"],
      "env": { "AGENTRAM_API_KEY": "agentram_your_key_here" }
    }
  }
}

Prefer the UI? Open Cursor Settings, go to Tools & MCP, click Add new MCP server, and paste the same block. If your install has an Enable MCP servers toggle, make sure it is on.

If you commit a project .cursor/mcp.json to version control, do not hardcode your key there. Keep the real key in a global file or an environment variable instead. Config paths and the settings layout can shift between Cursor versions; if the server does not appear, check Cursor's current MCP docs. The server command and env var stay the same.

Quick start

Reload the Cursor window so it picks up the server, then in the agent chat try:

Store my test command as "npm run test:ci" under the key test_cmd.

Later, in a new session:

Recall test_cmd from AgentRAM.

It reads the value back instead of starting from nothing. To see the temporal side, ask it to store a fact that changes, like a plan or a status, update it later, and then ask for the history.

Smoke test and troubleshooting

Give your agent memory

AgentRAM is a simple memory API for AI agents. One call to store, one to recall, shared across agents, no vector database. Store your first memory in about a minute.

Get your API key

1,000 free operations. No credit card.

© 2026 AgentRAM. All rights reserved.