Mem.ai MCP Server - Cloudflare Worker

A production-ready Cloudflare Worker that provides AI assistants with intelligent access to Mem.ai's knowledge management platform via the Model Context Protocol (MCP).

✨ Features

• 🧠 Intelligent Memory: Save and process content with Mem It's AI-powered organization
• 📝 Note Management: Create, read, and delete structured markdown notes
• 📁 Collections: Organize notes into searchable collections
• 🔒 OAuth Protected: GitHub OAuth authentication (optional)
• ⚡ Edge Deployment: Low-latency access via Cloudflare's global network
• 🆓 Free Tier: Runs on Cloudflare's generous free tier
• 🎯 6 MCP Tools: Complete Mem.ai API coverage

📋 Prerequisites

• Cloudflare account (free tier works!)
• Mem.ai account and API key
• Wrangler CLI installed
• Node.js 18+ and npm

🚀 Quick Start

Installation

# Clone the repository
git clone https://github.com/BurtTheCoder/mcp-mem.ai-cfw.git
cd mcp-mem.ai-cfw

# Install dependencies
npm install

# Run automated setup
./setup.sh

The setup script will:

1. Create KV namespaces (for OAuth mode)
2. Generate wrangler.toml configuration
3. Configure secrets (API keys)
4. Guide you through OAuth setup (if selected)

Deployment Modes

Option 1: OAuth-Protected (Recommended for Production)

Secure access with GitHub OAuth authentication.

# During setup, choose option 1
./setup.sh

# Deploy to Cloudflare
npm run deploy

OAuth Setup Requirements:

1. Create a GitHub OAuth App at https://github.com/settings/developers
2. Set callback URL to: https://mem-mcp-server.<your-subdomain>.workers.dev/github/callback
3. Configure Client ID and Secret during setup

Option 2: Simple Mode (Development)

No OAuth - direct API key authentication.

# During setup, choose option 2
./setup.sh

# Deploy to Cloudflare
npm run deploy

Local Development

# Start local development server
npm run dev

# Access at http://localhost:8787

🛠️ Available MCP Tools

1. mem_it - Intelligent Content Processing

Save and automatically process any content type with AI-powered organization.

Parameters:

• input (required): Content to save (text, HTML, markdown)
• instructions (optional): Processing instructions
• context (optional): Additional context
• timestamp (optional): ISO 8601 timestamp

Example:

{
  "input": "Meeting notes: Discussed Q1 roadmap...",
  "instructions": "Extract action items",
  "context": "Product Planning"
}

2. create_note - Create Structured Note

Create a markdown-formatted note with explicit organization.

Parameters:

• content (required): Markdown content
• collection_ids (optional): UUIDs of collections
• collection_titles (optional): Collection titles

3. read_note - Read Note by ID

Retrieve full note content and metadata.

Parameters:

• note_id (required): UUID of the note

4. delete_note - Delete Note

Permanently delete a note.

Parameters:

• note_id (required): UUID of the note

5. create_collection - Create Collection

Create a collection for organizing related notes.

Parameters:

• title (required): Collection title
• description (optional): Markdown description

6. delete_collection - Delete Collection

Delete a collection (notes remain, just unassociated).

Parameters:

• collection_id (required): UUID of the collection

⚙️ Configuration

Environment Variables (Secrets)

Set via wrangler secret put:

# Required
wrangler secret put MEM_API_KEY

# Optional (for OAuth mode)
wrangler secret put GITHUB_CLIENT_ID
wrangler secret put GITHUB_CLIENT_SECRET

wrangler.toml

Generated by setup.sh. Manual configuration:

name = "mem-mcp-server"
main = "src/github-oauth-index.ts"  # or src/simple-index.ts
compatibility_date = "2024-09-23"
compatibility_flags = ["nodejs_compat"]

[[durable_objects.bindings]]
name = "MCP_AGENT"
class_name = "MemMCPAgent"

[[migrations]]
tag = "v1"
new_sqlite_classes = ["MemMCPAgent"]

# For OAuth mode only:
[[kv_namespaces]]
binding = "OAUTH_KV"
id = "your_kv_namespace_id"
preview_id = "your_preview_kv_id"

🏗️ Architecture

src/
├── types.ts               # TypeScript interfaces & Zod schemas
├── mem-client.ts          # Mem.ai API client
├── mcp-agent.ts           # Durable Object with 6 MCP tools
├── github-oauth-index.ts  # OAuth-protected worker
└── simple-index.ts        # Simple worker (no OAuth)

Key Components

1. Durable Object (mcp-agent.ts)

• Stateful MCP server instance
• Handles all 6 Mem.ai tools
• Uses SQLite for persistence

2. API Client (mem-client.ts)

• TypeScript wrapper for Mem.ai API
• Error handling and validation
• Type-safe with Zod schemas

3. OAuth Handler (github-oauth-index.ts)

• GitHub OAuth flow
• PKCE support
• Token management with KV storage

4. Simple Handler (simple-index.ts)

• Direct SSE endpoint
• No authentication overhead
• Perfect for development

📚 API Endpoints

OAuth Mode

| Endpoint | Description | |----------|-------------| | / | Server information | | /authorize | OAuth authorization | | /token | Token exchange | | /github/callback | OAuth callback | | /sse | SSE endpoint (authenticated) | | /.well-known/oauth-authorization-server | OAuth discovery |

Simple Mode

| Endpoint | Description | |----------|-------------| | / | Server information | | /sse | SSE endpoint (no auth) | | /health | Health check |

🔧 Claude Desktop Integration

OAuth Mode

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "mem": {
      "url": "https://mem-mcp-server.<your-subdomain>.workers.dev/sse"
    }
  }
}

Claude will handle OAuth flow automatically.

Simple Mode

{
  "mcpServers": {
    "mem": {
      "url": "https://mem-mcp-server.<your-subdomain>.workers.dev/sse"
    }
  }
}

💰 Cloudflare Free Tier

This worker runs comfortably on Cloudflare's free tier:

• ✅ 100,000 requests/day
• ✅ 10ms CPU time per request
• ✅ Unlimited Durable Objects
• ✅ 1GB KV storage

Perfect for personal use and small teams!

🧪 Testing

# Test locally
npm run dev

# Check health endpoint
curl http://localhost:8787/health

# Test root endpoint
curl http://localhost:8787/

📊 Monitoring

View logs and analytics in Cloudflare Dashboard:

# View live logs
wrangler tail

# Check deployment status
wrangler deployments list

🐛 Troubleshooting

"Authentication failed"

• Check that MEM_API_KEY is set correctly
• Verify API key is valid at https://mem.ai

"Durable Object not found"

• Run migrations: wrangler migrations apply
• Redeploy: npm run deploy

OAuth callback fails

• Verify GitHub OAuth app callback URL matches your worker URL
• Check GITHUB_CLIENT_ID and GITHUB_CLIENT_SECRET are set

"KV namespace not found"

• Run setup.sh again to recreate namespaces
• Manually create via Cloudflare Dashboard

🤝 Contributing

Contributions welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Open a Pull Request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🔗 Links

• Mem.ai - Intelligent memory platform
• Mem.ai API Docs
• Model Context Protocol
• Cloudflare Workers
• Python Version - FastMCP Python implementation

💡 Related Projects

• mcp-limitless - Limitless AI MCP server (template for this project)
• mcp-mem.ai - Python FastMCP version

🙏 Acknowledgments

• Built with Cloudflare Workers
• OAuth flow based on mcp-limitless
• Powered by Mem.ai

Built with ❤️ using Cloudflare Workers and TypeScript