Skip to content

Guide · Agent skill

How to use the claude-mem-install skill for Cursor agents

Updated · written with AI from the project's own documentation and checked against our catalogue data

claude-mem-install is an agent skill that teaches an AI agent how to set up claude-mem, a memory plugin, inside the Cursor code editor. It is for developers who want persistent memory working in Cursor, and for anyone troubleshooting an existing claude-mem setup there. After this guide you will know what the skill does, how to add it to your agent, how to ask for it, and which setup option fits your situation.

Install claude-mem on Cursorthedotmack/claude-mem · Apache-2.0

Quick start

  1. 1Check Node.js. Confirm Node.js is installed and npx is available in your terminal, since the install command depends on it.
  2. 2Download the ZIP. Download the claude-mem-install skill ZIP from the directory page.
  3. 3Add the skill. Unzip it into ~/.claude/skills/ for personal use or the project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
  4. 4Choose a setup. Pick the local worker (default) or the remote worker with your own server URL.
  5. 5Run the install. Run npx claude-mem install --ide cursor for local, or add --runtime server and --server-url for remote.
  6. 6Configure the observer. Keep openrouter or cmem.ai for remote inference, or use --provider host with the shim on port 37778 for host-login inference.
  7. 7Set the token if remote. For a remote worker, set the CLAUDE_MEM_MCP_TOKEN plugin variable.

What does the claude-mem-install skill do?

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent reads when a task matches the skill's description. This skill covers installing claude-mem into Cursor. Its description says to use it when setting up claude-mem on Cursor, choosing between a local or remote worker, and choosing a local host-login observer or remote cmem.ai inference.

The skill covers two setups. The default is a local worker running on your own machine. The other is a remote worker on a server. It also explains how the observer, the part of claude-mem that summarises activity, gets its AI inference: either from a remote service at cmem.ai (through OpenRouter settings) or from a local host-login shim.

  • Local worker (default): runs on your machine.
  • Remote worker: runs on a server you specify, such as https://YOUR_HOST.
  • Observer inference: remote via cmem.ai/OpenRouter, or local via host login.
  • The skill is specific to Cursor and states that it does not require or cover other IDEs.

How do you install the skill?

Download the skill ZIP from the directory page. Then add it to your agent in one of two ways.

For Claude Code, unzip the file into your personal skills folder ~/.claude/skills/, or into a project's .claude/skills/ folder to limit it to that project.

In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.

  • Personal use: unzip into ~/.claude/skills/.
  • Single project: unzip into the project's .claude/skills/ folder.
  • Claude apps: Settings → Capabilities → Skills, then upload the ZIP.

How do you set up the local worker?

The default setup runs the claude-mem install command for Cursor. This starts the local worker. Hooks already ship with the plugin, and the MCP server named claude-mem-local talks to the worker.

For the observer, the typical choice is remote inference: leave the provider as openrouter or cmem.ai, with the base URL set to https://cmem.ai/api/inference/v1. If you prefer local host-login inference, use the --provider host option. That uses a loopback shim. Never bind the worker port. On macOS the worker is often on port 37777, so the shim uses 37778 instead.

  • Requires Node.js with npx available.
  • Remote inference: provider openrouter or cmem.ai.
  • Local host login: --provider host, shim on port 37778 (on macOS).
npx claude-mem install --ide cursor

How do you set up a remote worker?

Use a remote worker when you want a shared server rather than a process on your own machine. Replace YOUR_HOST with your server address. Then set the plugin variable CLAUDE_MEM_MCP_TOKEN so the claude-mem-remote MCP server can connect. The endpoint can be https://cmem.ai/api/mcp or your own server.

The skill also says never to restart a healthy worker, because it holds a RAM queue. Host and openrouter providers do not need the Claude CLI.

  • Set CLAUDE_MEM_MCP_TOKEN for the claude-mem-remote server.
  • Endpoint: https://cmem.ai/api/mcp or your own server.
  • Do not restart a healthy worker, since it holds a RAM queue.
npx claude-mem install --ide cursor --runtime server --server-url https://YOUR_HOST

How do you trigger the skill with a natural request?

You do not need to name the skill. Ask the agent in plain language about setting up claude-mem on Cursor, and it should match the request to this skill's description. Be specific about the setup you want, since the skill distinguishes between local and remote workers and between observer inference options.

If you are troubleshooting an existing install, say so and mention the problem, such as ports or provider settings. The skill includes those details.

  • "Set up claude-mem in Cursor with a local worker."
  • "Connect Cursor to a remote claude-mem server on my own host."
  • "My claude-mem observer in Cursor needs host-login inference. Which port should it use?"

What should you check for safety and privacy?

This skill runs scripts and uses the network. The default observer setup sends inference through cmem.ai or OpenRouter, and the remote worker option sends data to a server. If you do not want a background worker process or outside network calls, this skill is not a good fit. Review the install commands before the agent runs them, and check which provider and server URL are set.

The skill needs a host login for the local host-login observer option. Only enter the CLAUDE_MEM_MCP_TOKEN value if you intend to use the remote worker.

  • Scripts: yes. Network: yes.
  • Default inference leaves your data flowing to cmem.ai or OpenRouter.
  • Remote worker sends data to the server you specify.
  • Keep the worker port free; the shim uses 37778 on macOS.

What people use it for

  • Giving a Cursor coding agent persistent memory of your project through a local claude-mem worker.
  • Sharing one claude-mem server across a team by using the remote worker option.
  • Fixing an existing Cursor claude-mem install by checking ports and provider settings.
  • Choosing between cmem.ai/OpenRouter inference and local host-login inference for the observer.

Frequently asked questions

What is the claude-mem-install skill?

It is an agent skill that walks an AI agent through installing claude-mem, a memory plugin, into the Cursor code editor. It covers local and remote worker setups and observer inference options. It is a set of instructions the agent follows, not a standalone app.

How do I install claude-mem-install in Claude?

Download the skill ZIP from the directory page. Unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload the ZIP in the Claude apps under Settings → Capabilities → Skills.

What are the requirements for this skill?

You need Node.js with npx available, Cursor with the claude-mem plugin, and network access to cmem.ai or OpenRouter for the default remote inference. A remote worker also needs the CLAUDE_MEM_MCP_TOKEN plugin variable, and the host-login option needs a host login. The Claude CLI is not required for host or openrouter providers.

Is claude-mem-install free to use?

The source does not state pricing or licence terms for the skill's use beyond its Apache-2.0 licence. Check the claude-mem project and the cmem.ai service for any costs of the remote inference option.

Does this skill work with editors other than Cursor?

No. The skill is written for Cursor and states that it does not require or cover other IDEs or agents.

Should I use a local worker or a remote server?

Use the local worker if you want everything running on your own machine, which is the default. Choose the remote worker if your team needs a shared server, and remember it requires CLAUDE_MEM_MCP_TOKEN and sends data to the server you set.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits developers who want claude-mem memory working in Cursor, and teams choosing between a local worker and a shared server. It is less suitable if you want no background process or no outside network calls, or if you are not comfortable running terminal commands and editing environment variables. Users of other editors should look elsewhere.

Download the skill

Find exactly what
you're looking for