OpenMemory MCP is an open-source memory server that lets Claude, Cursor, and other LLM tools share context in real time. A live demo shows Claude saving personal information to a dashboard, then Cursor retrieving it instantly, with full access logs and controls to pause individual memories or clients.
Introducing OpenMemory MCP, a memory MCP server built to work across all your MCP clients. If you have used tools like Cursor or Claude, you have probably noticed one thing: they don't share context with each other. Everything resets. No continuity, no way to track or reuse what's already happened across tools. OpenMemory changes that. It's an open-source memory MCP server that lets your clients store and retrieve context in real time across sessions and across tools. You can run it easily by just running these two commands in your terminal. Now let me show you a demo. Once you run the commands, the dashboard is up and running. Using that, you can copy a simple command to install it in any of your MCP clients, including Claude Cursor or any other client. I have the Memory MCP server set up on Claude. I am going to ask Claude to remember things about me from my website. Just like that, Claude has saved things in memory. Now if I go back to the OpenAI OpenMemory dashboard and refresh, you can see all the memories getting populated. These are created by the source app, which is Claude. You also see memories automatically categorized. You can see all the apps registered on OpenMemory. Right now we have two apps registered. You can also see how many memories were created by a particular application and how many memories each MCP client has accessed, and you can go into details. Now let's go to Cursor and ask if it can recall things about me based on these memories. You can see it is able to fetch all the information about me that we saved from Claude. Back on the OpenMemory dashboard, Cursor now pops up. It shows that Cursor accessed 7 memories. You can check when certain memories were accessed by Cursor, how many times each memory was accessed, and you can pause access to certain memories or pause access for this MCP client altogether. Last but not least, everything stays local and nothing leaves your system. This is what shared memory for LLM apps should look like— fast, observable, and fully under your control. Check out OpenMemory. It is live, it is open source, and it is ready to plug into your stack.
