Quick Start
Install the MCP server, connect it to your agent, and store your first memory.
One command
Run this in your project directory:
npx memento-mcp init This signs you up, configures the MCP server, sets up Claude Code hooks, and creates .memento.json — all in one step. Restart Claude Code to load the new config.
Manual setup
If you prefer to configure things yourself, or aren't using Claude Code:
1. Sign up
Create a free account instantly — no email, no password, no OAuth:
curl -X POST https://memento-api.myrakrusemark.workers.dev/v1/auth/signup \
-H "Content-Type: application/json" \
-d '{"workspace": "my-project"}' Save the api_key from the response — you'll need it in the next step.
2. Configure your MCP client
Add this to your Claude Desktop, Claude Code, or any MCP-compatible client:
{
"mcpServers": {
"memento": {
"command": "npx",
"args": ["-y", "memento-mcp"],
"env": {
"MEMENTO_API_KEY": "mp_live_your_key_here",
"MEMENTO_API_URL": "https://memento-api.myrakrusemark.workers.dev",
"MEMENTO_WORKSPACE": "my-project"
}
}
}
} 3. Restart your client
The MCP server connects at startup. Restart so it picks up the new config.
First session
Verify the connection, store a memory, and recall it:
memento_health() // verify — shows workspace name, memory counts, status
memento_store(
content: "The user prefers dark mode and em dashes over parentheses.",
type: "instruction",
tags: ["preferences", "style"]
)
memento_recall(query: "user writing preferences") Recall combines keyword matching with vector similarity — searching "writing style" finds the memory above even though the exact words don't appear.
What to read next
- Core Concepts — understand the memory model
- The Protocol — orientation, recall hooks, writing discipline
- MCP Tools — full tool reference
- API Reference — every endpoint