Quick Start

Get AutoMem running in under 5 minutes

Step 1: Deploy AutoMem

Click the button below to deploy AutoMem to Railway. You'll need a Railway account (free to sign up).

Cost: Railway charges approximately $5/month for AutoMem (includes FalkorDB + memory API service). Qdrant Cloud offers a free tier for the vector database.

What Railway Will Set Up:

FalkorDB instance (graph database)
AutoMem API service with automatic health checks
SSL certificate for secure connections
Environment variables pre-configured (you'll add Qdrant credentials in Step 3)

Note: You should see exactly 2 services in Railway: memory-service and falkordb

Step 2: Get Your Credentials

After deployment completes:

Go to your Railway project dashboard
Click on the memory-service service
Go to the Variables tab
Find and copy these values:
- AUTOMEM_API_KEY
- OPENAI_API_KEY (you'll set this)
Note your AutoMem endpoint URL (shown in the Settings tab)

Tip: Your endpoint URL will look like https://memory-service-production-XXXX.up.railway.app

Step 3: Set up Qdrant Cloud

Qdrant is the vector database that powers AutoMem's semantic search. It's hosted separately on Qdrant Cloud's free tier.

Create a free Qdrant Cloud account:
- Go to https://cloud.qdrant.io/
- Sign up (it's free, no credit card required)
Create a cluster:
- Click "Create Cluster"
- Choose the Free tier (1GB storage, perfect for personal use)
- Select a region close to your Railway deployment
- Wait for the cluster to be provisioned (~2 minutes)
Get your credentials:
- Click on your cluster name
- Copy the Cluster URL (looks like https://xxxxx-xxxxx.aws.cloud.qdrant.io:6333)
- Copy the API Key from the "Access" tab
Add credentials to Railway:
- Go back to your Railway project
- Click on the memory-service
- Go to Variables tab
- Add these two variables:
  
  QDRANT_URL=https://xxxxx-xxxxx.aws.cloud.qdrant.io:6333
  
  QDRANT_API_KEY=your-api-key-here
- The service will automatically restart with the new configuration

Success! Your AutoMem service is now connected to both FalkorDB (graph) and Qdrant (vector) databases.

Step 4: Configure Your MCP Client

Choose your AI assistant:

Claude Desktop Configuration

Edit ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "automem": {
      "command": "npx",
      "args": [
        "-y",
        "@verygoodplugins/mcp-automem@latest"
      ],
      "env": {
        "AUTOMEM_ENDPOINT": "https://your-railway-url.up.railway.app",
        "AUTOMEM_API_KEY": "your-api-key-here"
      }
    }
  }
}

Replace your-railway-url and your-api-key-here with your actual values. Then restart Claude Desktop.

Cursor IDE Configuration

Alternatively, add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "automem": {
      "command": "npx",
      "args": [
        "-y",
        "@verygoodplugins/mcp-automem@latest"
      ],
      "env": {
        "AUTOMEM_ENDPOINT": "https://your-railway-url.up.railway.app",
        "AUTOMEM_API_KEY": "your-api-key-here"
      }
    }
  }
}

Replace the placeholders and restart Cursor. The MCP tools will appear in the tools panel.

Step 5: Test It Out

Try these commands in your AI assistant:

Store a memory:

"Remember that I prefer TypeScript over JavaScript for new projects"

Recall memories:

"What programming language should I use for my new project?"

Check memory health:

"Check the status of my AutoMem service"

Success! Your AI now has persistent memory across all devices and conversations. Memories are automatically synced and available everywhere.

Troubleshooting

I only see 2 services in Railway, where's Qdrant?

This is correct! Qdrant is hosted separately on Qdrant Cloud, not on Railway. You should see exactly 2 services:

memory-service - The AutoMem API
falkordb - The graph database

Make sure you completed Step 3 to set up Qdrant Cloud and add the credentials to Railway.

Health check is failing

Check that all environment variables are set correctly:

QDRANT_URL - Should include the full URL with port (e.g., https://xxxxx.cloud.qdrant.io:6333)
QDRANT_API_KEY - Your Qdrant cluster API key
FALKORDB_URL - Auto-configured by Railway

After updating variables, wait 1-2 minutes for the service to restart.

MCP tools not appearing in Claude/Cursor

Try these steps:

Verify your AUTOMEM_ENDPOINT doesn't have a trailing slash
Check that AUTOMEM_API_KEY is copied correctly (no extra spaces)
Restart Claude Desktop or Cursor completely (not just reload)
Check the MCP logs for connection errors

Next Steps

Explore Features →

Learn about memory types, knowledge graphs, and advanced capabilities

View Source Code →

Contribute to AutoMem or run it locally for development