Skip to content

meldfunction/kitchen-table

Repository files navigation

🍳 Kitchen Table Cloud

Bring tech back to the kitchen table.

Community cooperative infrastructure — AI, files, music, messaging, and more — owned, governed, and maintained by the people who use it. No venture capital. No algorithm. No rent-seeking.

License: MIT Community: Open Hardware: Raspberry Pi 5


In Three Commands

# Install the AI runtime
curl -fsSL https://ollama.com/install.sh | sh

# Pull a capable free model (2.4GB — less than one month of ChatGPT Plus)
ollama pull phi3.5:3.8b-mini-instruct-q4_K_M

# Talk to it. On your hardware. Privately. Forever.
ollama run phi3.5:3.8b-mini-instruct-q4_K_M "Hello. Are you mine now?"
# It says: Yes. Completely.

Cancel a subscription. Come back for the rest.


What This Is

Every major tech platform — Spotify, Google Drive, ChatGPT, Discord — is infrastructure you depend on but don't control. The prices go up. The terms change. Your data trains their models. You have no vote.

Kitchen Table Cloud is a blueprint and community for building the alternative: community-owned, cooperatively governed digital infrastructure that's as capable as the commercial stack but ruled by its users, not shareholders.

The full technical guide — 3,000+ lines, seven parts, full appendices — is in sovereign-stack-guide.md. Check out the read me as well.

The human-readable version is at Kitchen Table.


What You Can Run

Category Corporate Version Community Version Difficulty
AI Assistant ChatGPT, Claude Ollama + open models Easy
Distributed AI GPT-4 API cluster Petals + 6× Pi 5 nodes Medium
File Storage Google Drive NextCloud Medium
Music Spotify Navidrome Easy
Photos Google Photos Immich Medium
Video Library Netflix, Plex Jellyfin Easy
Chat Discord, WhatsApp Matrix + Element Medium
Smart Home Nest, Alexa Home Assistant Medium
Search Google SearXNG Easy (30 min)
Passwords 1Password, LastPass Vaultwarden Easy
DNS + Ad Blocking Google 8.8.8.8 Pi-hole + Unbound Medium
Off-Grid Comms Cell signal Meshtastic LoRa Medium
Code Hosting GitHub (Microsoft) Gitea Medium
Home Dashboard CasaOS Easy

Repository Structure

kitchen-table-cloud/
├── sovereign-stack-guide.md  ← THE GUIDE — full technical walkthrough (Parts 1–7 + Appendix)
├── stitch_guide.py           ← Python build script that assembles the guide from source files
├── docs/
│   ├── hardware.md           ← Pi 5 specs, NVMe boot, cluster builds
│   ├── inference-stack.md    ← Ollama, llama.cpp, Petals setup in detail
│   ├── open-models.md        ← Model selection, licenses, fine-tuning
│   ├── self-hosted-services.md ← Full service configs: NextCloud, Matrix, Navidrome...
│   ├── mesh-network.md       ← Tailscale ACLs, Ansible, Yggdrasil
│   └── exit-strategy.md      ← Phase-by-phase provider independence roadmap
├── config/
│   ├── openclaw-base-config.json
│   ├── tailscale-acls.json
│   └── ollama-config.json
├── playbooks/
│   ├── onboard-new-member.yml  ← Ansible: full Pi setup from scratch
│   ├── update-all.yml          ← Push updates to all nodes simultaneously
│   └── deploy-services.yml     ← Deploy full service stack
├── source-skills/
│   ├── open-claw/              ← OpenClaw configuration skill
│   └── open-claw-net/          ← Distributed Pi cluster skill
└── index.html                  ← GitHub Pages main site

Guide Structure

The sovereign-stack-guide.md covers everything:

Part Topic What You Get
1 OpenClaw basics CLAUDE.md, permissions, sessions, subagents, hooks, MCP
2 Advanced configuration Deep CLAUDE.md patterns, hook recipes
3 Sovereign stack Pi hardware, OS, Ollama, llama.cpp, Petals, gateway config
4 Self-hosted services NextCloud, Navidrome, Matrix, SearXNG, Gitea full configs
5 Open models Which to use, licenses, fine-tuning on group data
6 Mesh networking Tailscale ACLs, Ansible, Yggdrasil deep dive
7 Inspiration + references Communities, projects, full reading list
Appendix Ready-to-run Quick-start checklist, cost breakdown, security checklist, migration

Hardware

Per person: Raspberry Pi 5 (8GB) + 512GB NVMe SSD + M.2 HAT+ + active cooler + 27W PSU ≈ ~$160 all-in

What a group can do together:

Group Size Collective RAM AI Capability
1 person 8GB Phi-3.5 Mini, Llama 3.2 3B — handles 80% of everyday tasks
3 people 24GB Llama 3.1 8B at full precision via Petals
6 people 48GB Llama 3.1 8B + Mistral 7B simultaneously
10 people 80GB Qwen2.5 14B Q4 — approaching frontier capability

Ongoing cost: ~$4/month electricity per node vs $67–109/month in subscriptions replaced.


Security

Read security.html before opening access to your community.

Key points:

  • Gateway must bind to Tailscale IP only — never 0.0.0.0 (CVE-2026-25253)
  • Skills locked to verified sources — ClawHavoc campaign compromised 1,400+ unverified skills
  • Run openclaw doctor before connecting any channels
  • SOUL.md per user — explicit "never do this" boundaries for every member's agent

Get Involved

This is a cooperative, not a startup. Five ways to participate:

  1. Supporter — Free. Tell people this exists. Share it. Show up.
  2. Member — Pay-what-you-can. Access all community-run services.
  3. Node Runner — Contribute hardware. Your Pi = community infrastructure. Electricity stipend.
  4. Builder — Contribute skills: tech, writing, teaching, design. Your expertise is infrastructure.
  5. Steward — Governance and funding. Shape decisions. Sponsor nodes for members who can't afford hardware.

To join: Open an issue titled "I want to get involved" and tell us what you're working with.


Philosophy

We believe:

  • Tech should connect, not entrap. The extractive model is a design choice, not a law of nature.
  • Cooperative governance produces better infrastructure than corporate governance — for the people who use it.
  • The window is open. Open models improve every 3–6 months. Hardware is cheap. Build now.
  • Local-first scales. What works for one person's notes (Obsidian) works for a community's infrastructure.

We're inspired by Obsidian's local-first philosophy, Anytype's privacy-as-architecture approach, Elicit's honesty about AI capabilities, the Calm Tech Institute's attention-respecting design principles, and Patagonia's mission-as-governance model.

Full vision: vision.html · Full map: map.html


Contributing

PRs welcome. Open an issue before large changes to discuss direction.

Most valuable contributions right now:

  • Setup guides for specific configurations (Windows host, macOS, specific Pi variants)
  • Translations of getting-started documentation
  • Corrections to any technical claims that are wrong or outdated
  • Real-world configs from groups that have deployed this

See CONTRIBUTING.md.


License

MIT. Use it. Build on it. Share it. Make it better.


Built April 2026. Open source. No ads. No VC. No quarterly earnings call. Community governed · Local-first · The time is now.

About

No description, website, or topics provided.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages