Give Claude Code, Cursor and ChatGPT one shared memory
Give Claude Code, Cursor and ChatGPT one shared memory If you use more than one AI, you know the tax: every new tool starts from zero. You explain the project, your preferences, your rules β again. Switch tools mid-job
Give Claude Code, Cursor and ChatGPT one shared memory
If you use more than one AI, you know the tax: every new tool starts from zero. You explain the project, your preferences, your rules β again. Switch tools mid-job and the next AI has no idea where the last one stopped.
OwlMem fixes that with one shared memory. (Disclosure: OwlMem is our own product β owlmem.com.) Tell one AI something once, and every connected AI can recall it. What one learns, they all know.
The setup: one address, about a minute per AI
Every client connects to the same MCP server:
https://mcp.owlmem.com/mcp
The fastest way is to let your AI do it. In Claude Code, Codex, Cursor, or any AI that can run commands, say:
Read https://mcp.owlmem.com and set up OwlMem for yourself.
It reads the setup guide, runs the installer, and asks you to sign in. All you do is click Allow. Or run the installer yourself:
npx -y owlmem install
It finds the AI apps on your computer and adds OwlMem to each one, merging into your existing config files instead of overwriting them, and backing up anything it changes. I ran it in dry-run mode to see what it would do on a machine with Codex installed:
OwlMem install β connects the AI apps on this computer to https://mcp.owlmem.com/mcp (dry run, nothing is changed)
β Claude Code not installed
+ Codex will add OwlMem (~/.codex/config.toml)
will write the OwlMem rules block in ~/.codex/AGENTS.md
β Cursor not installed
β VS Code not installed
β Windsurf not installed
β Gemini CLI not installed
β Claude Desktop not installed
Dry run: 1 app(s) would change. Run again without --dry-run to apply.
If you'd rather add each one by hand, the commands are short:
Claude Code:
claude mcp add --transport http --scope user owlmem https://mcp.owlmem.com/mcp
Then start Claude Code, type /mcp, select owlmem, and choose Authenticate.
Cursor: add the URL to ~/.cursor/mcp.json, then open Cursor Settings β MCP. OwlMem shows Needs login β click it and approve in the browser.
{
"mcpServers": {
"owlmem": { "url": "https://mcp.owlmem.com/mcp" }
}
}
ChatGPT: open Settings β Apps & Connectors, turn on Developer mode under Advanced settings, choose Create, name it OwlMem, paste the URL, and choose OAuth. One caveat from the docs: availability of custom connectors depends on your ChatGPT plan and workspace settings β if you don't see the option, your plan may not include it.
Codex:
codex mcp add owlmem --url https://mcp.owlmem.com/mcp
Whichever way you connect, the first connection opens an OwlMem page: sign in with your email, name the AI, click Allow. No keys to copy. Every connected AI shows up in the console under Agents & tools β Agents, where you can switch it off or delete it any time.
Try it: teach one, ask another
Once two AIs are connected, the test takes thirty seconds:
- In Claude Code, say: Record the project I'm working on in OwlMem, so my other AIs know it too.
- In ChatGPT, ask: How is this project deployed?
ChatGPT answers from what Claude recorded β no copy-pasting context between apps. That's the whole point: the memory lives in one place you control, not inside any single AI's chat history. Switch tools whenever you like and keep it all.
What it costs
Free for two AIs, forever β no card. That covers the exact setup above: one coding AI and one chat AI sharing a memory. Paid plans add more AIs, storage, and memory writes; every plan gets every feature.
If you already juggle Claude Code, Cursor, and ChatGPT, the shared memory is the part you'll feel first: you stop paying the re-explaining tax on every switch. Set it up from the docs β about a minute per AI.
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes β full credit and traffic to the original publisher.
