marcotte-dev

xmmarcotte/marcotte-dev

3.2

If you are the rightful owner of marcotte-dev and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.

The mcp-server-qdrant is a Model Context Protocol server designed to integrate with the Qdrant vector search engine, providing a semantic memory layer for storing and retrieving information.

Tools
2
Resources
0
Prompts
0

marcotte-dev

Infrastructure and services for my Oracle Cloud Always Free instance.

Tailscale IP: <TAILSCALE_IP> (private mesh network - see GitHub Secrets) Public IP: <PUBLIC_IP> (Oracle Cloud - see GitHub Secrets) Domain: marcotte.dev (future)

Services

Semantic memory and codebase intelligence MCP server for Cursor IDE.

Status: ✅ Production (deployed and running) Endpoint: http://<TAILSCALE_IP>:3856/mcp Tech: Python, FastMCP, Qdrant, FastEmbed

Features:

  • Persistent memory across all machines
  • Semantic memory search across all projects
  • Architectural decision tracking
  • Code pattern recognition and storage
  • Pure memory-first workflow (no codebase indexing)
  • Cursor integration for local code intelligence

See for details.

Infrastructure

Oracle Cloud Always Free

  • Instance: VM.Standard.A1.Flex
  • CPU: 4 ARM cores (Ampere Altra)
  • RAM: 24GB
  • Storage: 200GB boot volume
  • Cost: $0/month (free forever)

Networking

  • Tailscale: Private mesh VPN for secure access
  • Public IP: Only for initial setup, then Tailscale-only
  • Firewall: iptables + Oracle Cloud security lists

Backups

  • Primary: Automated daily rsync to local machine via Tailscale
  • Location: ~/.marcotte-dev-backup/ on your laptop/desktop
  • Schedule: Daily at 2:00 AM (with catch-up on boot if missed)
  • Method: Systemd timer (runs on your local machine)
  • Retention: Single current mirror (fast recovery)
  • Scripts: See

Quick Start

🎯 Option 1: GitHub Actions CI/CD (Manual Trigger)

Deploy from GitHub Actions:

  1. Set up GitHub Secrets (see )
  2. Go to ActionsDeploy to Oracle CloudRun workflow
  3. Choose action and click Run workflow
  4. GitHub Actions handles everything automatically!

Benefits:

  • ✅ No local setup required
  • ✅ Data preservation during redeploys
  • ✅ Full infrastructure automation
  • ✅ Run on-demand when you're ready

See for details.

🚀 Option 2: Local Automated Provisioning

First time setup:

  1. Configure Terraform (see )
  2. Run full provisioning:
    ./scripts/provision.sh
    
    This automatically:
    • Provisions infrastructure (VM, networking, security)
    • Installs Docker, Tailscale, and system updates
    • Restores data from backup (if exists)
    • Deploys all services

Redeploy after updates:

# Just run the same command - it handles everything including data preservation!
./scripts/provision.sh

Manual Deployment (If Instance Already Exists)

# Deploy all services
./scripts/deploy.sh <oracle-tailscale-ip>

# Deploy specific service
./scripts/deploy.sh <oracle-tailscale-ip> spot-mcp-server

Set Up Backups

Run on your local machine (laptop/desktop):

# Systemd timer - runs daily at 2 AM, catches up on boot if missed
./scripts/setup-systemd-backup.sh <TAILSCALE_IP> ~/.marcotte-dev-backup

Why this works:

  • ✅ Runs missed backups when you boot up (laptop was off at 2 AM? No problem!)
  • ✅ Prevents duplicate runs if you reboot multiple times
  • ✅ Pulls data from Oracle instance via Tailscale (secure)
  • ✅ Easy to monitor with systemctl --user list-timers

Monitoring

# SSH to instance (via Tailscale - recommended)
ssh ubuntu@<TAILSCALE_IP>

# Or via public IP
ssh ubuntu@<PUBLIC_IP>

# Check all services
docker ps

# Check specific service
docker logs -f spot-mcp-server

# Check resources
htop
df -h

Documentation

  • - How to use Spot with Cursor IDE
  • - Automated deployments from GitHub
  • - Quick guide to configure CI/CD secrets
  • - System design and components
  • - Infrastructure as Code

Adding a New Service

  1. Create services/your-service/ directory
  2. Add Dockerfile and service code
  3. Add to services/docker-compose.yml if it should orchestrate with others
  4. Update this README
  5. Test locally, then deploy with ./scripts/deploy.sh

Current Status

Instance: ✅ Deployed and running Services: ✅ Spot MCP Server operational Backups: ✅ Automated daily backups configured Cost: $0/month (Always Free tier)

Oracle Cloud Always Free includes:

  • 2 VM.Standard.A1.Flex instances (4 OCPUs, 24GB RAM total)
  • 200GB boot volumes
  • 10TB/month outbound data transfer
  • Permanent (not trial-based)

Future Plans

  • Host at marcotte.dev domain
  • Add monitoring dashboard (Grafana?)
  • Add more services as needed
  • Consider multi-region backup

License

Apache 2.0