MemoryLayer stores your conversations and retrieves relevant context automatically — so your AI never forgets who you are.
Every time you open a new chat with Claude — it has no idea who you are, what you're working on, or what you discussed yesterday. You repeat yourself, constantly.
Tell Claude "save this to memory." MemoryLayer extracts key facts using LLaMA 3.3 70B and stores them as searchable vectors.
Ask anything next session. MemoryLayer finds the most relevant memories and injects them automatically.
Your AI now knows your projects, preferences, and history — across every session.
Conversations distilled into specific searchable facts using LLaMA 3.3 70B
Memories retrieved by meaning not keywords
Recent memories rank higher, old ones fade
Every user has isolated memory space via Supabase
Similarity above 90% is detected and skipped
Works with Claude Desktop via MCP protocol out of the box
Sign up at the bottom of this page. Remember your email and password — you will need to manually enter them in the Claude Desktop config file in Step 4. There is no dashboard or login page.
Download mcp_server.py from the GitHub repo and save it anywhere on your computer. Note the full file path — you will need it in Step 4.
Open the config file in any text editor. The location depends on your OS:
~/Library/Application Support/Claude/claude_desktop_config.json. To open the folder run: open ~/Library/Application\ Support/Claude/ in Terminal.
%APPDATA%\Claude\ and hit Enter. Open claude_desktop_config.json.
%LOCALAPPDATA%\Packages\Claude_pzs8sxrjxfjjc\LocalCache\Roaming\Claude\ and hit Enter. Open claude_desktop_config.json.
⚠️ Microsoft Store installs have two config files. Claude reads from LocalCache\Roaming, NOT from %APPDATA%\Claude. Use the path above.
Paste this into your claude_desktop_config.json. Replace /path/to/mcp_server.py with the actual full path of the file you downloaded in Step 2.Paste the full python path in "command".
To find the python path write in terminal which python this will give you the full python path
{
"mcpServers": {
"Memory-Layer": {
"command": "python",
"args": ["/path/to/mcp_server.py"],
"env": {
"MEMORY_EMAIL": "your@email.com",
"MEMORY_PASSWORD": "yourpassword",
"RAILWAY_URL": "https://memory-production-d533.up.railway.app"
}
}
}
}Replace /path/to/mcp_server.py with the actual full file path where you saved mcp_server.py in Step 2. For example: /Users/yourname/Documents/mcp_server.py on Mac or C:\Users\yourname\Documents\mcp_server.py on Windows.
You must install dependencies first by running: pip install fastmcp requests python-dotenv
Fully quit and reopen Claude Desktop — don't just close the window.
The Memory-Layer tools will appear in your conversation toolbar after restart.
Save this conversation to memory or Save this to the memory-layer MCPWhat do you know about me? Load from memory.Sign up with your email and password. Remember what you enter here — you will manually type this email and password into the Claude Desktop config file in Step 4 of the setup guide.