Setting up memory, three clicks, no terminal
Create an account, click Connect, confirm OAuth. In five minutes your AI tool has persistent memory.
What you need
- An AI tool that supports MCP: Claude Desktop, Claude Code, Cursor, or Codex
- A browser
- A Google, GitHub, or Discord account, or an email address for Magic Link
No terminal, no API key copying, no server to set up. We use the SaaS option. Self-hosting is mentioned further down as an option for advanced users.
Step 1: Create an account
Go to memory.studiomeyer.io and click "Sign in". You have four ways:
- With Google
- With GitHub
- With Discord
- With Magic Link to your email
Each is equivalent. No password, no credit card. You land directly in the dashboard.
Step 2: Connect your AI tool
In the dashboard you'll see a button "Connect MCP Server" at the top. Click it. A list of supported clients appears: Claude Desktop, Claude Code, Cursor, Codex.
Pick your tool. You get a Connect link. Clicking it does the following:
- Your AI tool opens (or asks for permission to open).
- It starts an OAuth flow back to memory.studiomeyer.io.
- You confirm in the browser that the tool can have access.
- Done. Memory is connected.
In Claude Desktop a small hammer icon appears in the chat bar. Click it and you'll see the memory tools.
In Claude Code type /mcp and you'll see the memory server in the list.
In Cursor and Codex it's the same, just in their respective tool panels.
Step 3: First test
In your AI tool type:
Remember: I'm working on a project called "Meeting App". Stack is Next.js 16 and Prisma.
The AI replies that it remembered. In the background it called a tool that wrote the entry to the database.
Close the session. Wait briefly. Open a new session.
Now type:
What do you know about my project?
The AI searches the memory and says: "I know you're working on the Meeting App, Next.js 16 and Prisma." That's persistent memory.
What's happening behind the scenes
Your first sentence was saved as a learning entry. On the second sentence the AI searched the database both full-text and semantically, found the relevant entries, and loaded them into context before answering.
Every new session no longer starts from zero. The longer you use the system, the more it learns about you, your projects, your workflow.
When something doesn't work
Connection fails. Check that your AI tool is on the latest version. MCP support landed in Claude Desktop in early 2025, Claude Code in mid-2025, Cursor and Codex in late 2025.
Memory tools aren't visible. Restart the tool. Claude Desktop benefits from "Quit and Reopen", Claude Code only needs a new session.
OAuth browser doesn't open. As a backup the dashboard at memory.studiomeyer.io exposes an API key. You can paste it manually into your tool's MCP configuration. That's the fallback path, the Connect button is the normal one.
Local memory as an alternative
If you don't want your data in a cloud SaaS: there's an open-source variant called local-memory-mcp. It runs as a small process on your own machine. SQLite database, everything stays local.
Pros:
- Fully private, no data leaves your machine
- Free, no SaaS fee
- You control everything yourself
Cons:
- Setup is more technical, terminal, npm install, configuration
- Available only on one machine, your memory isn't synced between laptop and desktop
- No knowledge graph, no multi-agent, no automatic decay
- You maintain it: backups, migrations, updates
- No web dashboard
If you're the type who self-hosts everything: Level 6 shows you how to build and run the memory server yourself. Until then the SaaS path is fine for learning.
That's also the honest promise of this academy: we show you the fast SaaS path so you can learn how memory works. If you decide after the learning phase that self-hosting fits you better, you'll have the knowledge to do it. Both are legitimate.
What's next
In Lesson 4 you learn Hooks and Skills, the mechanisms that let memory run automatically. Instead of typing "remember" you'll configure the AI to save entries on its own when it makes sense.