| Feature | Description |
|---|---|
| Model Context Protocol | Native Model Context Protocol support for Claude, Cursor, and other AI clients |
| Multi-Platform | Deploy on Mac, Raspberry Pi, or Cloud VPS with one command |
| Infrastructure as Code | Production-ready Ansible Playbooks, Helm Charts, Docker Compose |
| Observability | Built-in SQLite activity metrics, beautiful time-series charts, audit logs |
| Ansible-First Prod | Senior DevOps Workflow: Automated deployment, health-checks, and state management |
| Hybrid CI/CD | Self-Hosted RPi Runner for native 64-bit ARM builds + GitHub Cloud for x86 |
| Security | Traefik reverse proxy, Tailscale Funnel for zero-config HTTPS |
To install CyberMem on your local machine, run:
npx @cybermem/cli installand follow the instructions in terminal.
Full Quick Start guide for every platform is available at cybermem.dev/#quickstart.
Problem: Your AI tools (Claude, Cursor, Antigravity) don't share memory. Each session starts fresh.
Solution: CyberMem gives them a shared, persistent memory layer.
| Without CyberMem | With CyberMem |
|---|---|
| Claude forgets your project context | All tools remember your preferences |
| Cursor doesn't know your coding style | Context persists across sessions |
| Each tool has separate knowledge | One unified memory for all AI agents |
For Platform Engineers: CyberMem demonstrates advanced Infrastructure practices:
- IaC Automation: CLI generates Docker Compose, Ansible, or Helm depending on the target.
- Hybrid CI: Leverages a private Raspberry Pi runner to bypass QEMU overhead, achieving native ARM64 build speeds.
- Zero-Trust Access: Integrates Tailscale Funnel for secure, public access without port forwarding.
CyberMem is a reference implementation of production MCP patterns:
🔐 Multi-tenant auth: Traefik ForwardAuth extracts MCP client identity into audit logs
📊 Zero-overhead observability: Vector pipeline converts logs → Prometheus metrics
🚀 Multi-platform IaC: CLI auto-generates Docker Compose / Ansible / Helm per target
☸️ Production deployment: Kubernetes with rolling updates, health checks, PVC
🔧 Hybrid CI/CD: Self-hosted ARM64 runner bypasses QEMU 10x slowdown
551+ commits • 25 releases • Full E2E testing
If you're building MCP infrastructure at companies like Descope, Anthropic, or similar — use this as a production pattern library.
📧 Open to MCP/Agentic Infrastructure roles: mikhailkogan17@gmail.com
---
config:
theme: base
wrappingWidth: 290
fontSize: 11
---
graph TD
subgraph CI["🚀 Hybrid CI/CD Pipeline"]
GHA["**GitHub Actions**<br/>(Control Plane)"]
Cloud["**GitHub Cloud**<br/>(AMD64 Builds)"]
RPiR["**Self-Hosted RPi**<br/>(Native ARM64 Builds)"]
GHA --> Cloud
GHA --> RPiR
end
subgraph Gen["🏗️ IaC Engine"]
CLI["**CLI**"]
Templates["**Infrastructure Templates**"]
CLI --> Templates
Compose["**Docker Compose**<br/>(Local)"]
Ansible["**Ansible**<br/>(IoT/Edge)"]
Helm["**Helm Charts**<br/>(Cloud/K8s)"]
Templates --> Compose
Templates --> Ansible
Templates --> Helm
end
subgraph Runtime["⚙️ CyberMem Runtime"]
Traefik["**Traefik**<br/>(Reverse Proxy)"]
LE["**Log Exporter**<br/>(Access Logs)"]
DBE["**DB Exporter**<br/>(SQLite Metrics)"]
Dash["**Dashboard**<br/>(Monitoring UI)"]
OM["**Core API**<br/>(Memory Engine)"]
DB["**SQLite / Postgres**<br/>(Persistence)"]
Traefik -->|Logs| LE
Traefik -->|API| OM
OM --> DB
LE --> DB
DB --> DBE
DBE --> Dash
end
Cloud -.-> Runtime
RPiR -.-> Runtime
Compose -.-> Traefik
Ansible -.-> Traefik
Helm -.-> Traefik
CyberMem CLI provides a standardized set of commands for complete lifecycle management:
npx @cybermem/cli install # Install/Initialize services (Mac/RPi/VPS)
npx @cybermem/cli uninstall # Stop and teardown services
npx @cybermem/cli upgrade # Pull latest images and update instance
npx @cybermem/cli backup # Create a data backup (.tar.gz)
npx @cybermem/cli restore # Restore from a backup file
npx @cybermem/cli reset # Wipe database (Destructive!)
npx @cybermem/cli dashboard # Open monitoring dashboardImportant
Ansible-First Prod: For Raspberry Pi or remote servers, the CLI automatically leverages Ansible to ensure state-consistent, safe, and verifiable deployments.
cybermem/
├── packages/
│ ├── cli/ # Command-line tool (TypeScript)
│ │ ├── src/ # CLI logic
│ │ └── templates/ # ⭐ Infrastructure templates
│ ├── mcp/ # MCP Server & Core Engine (TypeScript)
│ │ └── src/ # Tooling & Memory Logic
│ └── dashboard/ # Monitoring UI (Next.js)
├── docs/ # Documentation sources
├── .github/
│ └── workflows/ # ⭐ CI/CD pipelines
└── README.md
Key innovation: packages/cli/templates/ contains the infrastructure-as-code templates.
The CLI reads these, interpolates variables, and generates production configs.
Full documentation available at docs.cybermem.dev:
| Guide | Description |
|---|---|
| Local Setup | Mac/Linux development environment |
| Ansible Deployment | Production standard for RPi/Edge |
| Raspberry Pi | Edge deployment with Tailscale |
| Cloud/VPS | Production Kubernetes deployment |
| MCP Integration | Connect Claude, Cursor, and more |
Standard Node.js auth middlewares often fail on underpowered Edge devices (RPi) or cause high latency. CyberMem uses Traefik as a Reverse Proxy to handle authentication at the networking layer. This allows the Core API to remain "clean" and deterministic, while Traefik extracts identity headers (X-Client-Name) into audit logs before the request even hits the application.
We follow the Infrastructure Appropriateness principle.
- RPi/Edge: Needs mutable state management and OS-level hardening (docker-compose, systemd). Ansible ensures idempotent state without the overhead of a control plane.
- Cloud/VPS: Scaling and high availability are paramount. Helm allows us to leverage Kubernetes native primitives (Ingress, PVC, HPA) for a truly elastic platform.
We use tools/test-k8s.sh and the CyberMem Gatekeeper to guarantee that every release is stable. Below is the raw console verification of a production-grade deployment.
NAMESPACE: cybermem
NAME READY STATUS RESTARTS AGE
pod/cybermem-dashboard-6dd67f5586-djwwh 1/1 Running 0 2m
pod/cybermem-openmemory-65fdf6d85c-g628g 1/1 Running 0 2m
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
cybermem-lb LoadBalancer 10.43.0.1 127.0.0.1 8626/TCP 2m
cybermem-dashboard ClusterIP 10.43.79.62 <none> 3000/TCP 2m
cybermem-openmemory ClusterIP 10.43.95.212 <none> 8080/TCP 2m
# Proof of zero-drift state management on Raspberry Pi
ansible-playbook -i inventory/hosts.ini playbooks/deploy-cybermem.yml
PLAY [Deploy CyberMem to Raspberry Pi] **
TASK [Gathering Facts] **
ok: [raspberrypi.local]
TASK [cybermem : Pull latest images from GHCR] **
ok: [raspberrypi.local] => (changed=false)
TASK [cybermem : Start services] **
ok: [raspberrypi.local] => (changed=false)
PLAY RECAP **
raspberrypi.local : ok=15 changed=0 unreachable=0 failed=0 skipped=0Contributions are welcome! See CONTRIBUTING.md for development setup and guidelines.
MIT
Created by Mikhail Kogan
📧 Open to MCP/Agentic Infrastructure roles: mikhailkogan17@gmail.com
🔗 LinkedIn • CV (PDF)
