Self-host · the watch is yours

The Quilt is yours if you want it. Three paths, from "play with it on my laptop" to "deploy the whole ecosystem to my own domain." Pick what you can carry.

01Which path is right for you?

Path	Time	What you get	What you need
Local Python	5 min	Run the substrate on your laptop. Run the tests. Modify the apps. No network.	Python 3.10+, git
Cloudflare Worker	15 min	Deploy the LLM proxy to your own Cloudflare account. Get a free rate-limited LLM endpoint.	Cloudflare account, wrangler
Full Stack	1 day	Clone the whole ecosystem. Deploy to your own domain. Vectorize your own canon. Save state in your own KV.	Cloudflare account, GitHub, ~1 hour of reading

02Path 1: Local Python

Run the original Python substrate. 405 tests. 5 apps. Modify anything you want.

git clone https://github.com/SuperInstance/quilt-substrate.git
cd quilt-substrate
python3 -m venv .venv && source .venv/bin/activate
pip install -e .
python3 -m pytest tests/ -q
# 405 passed in 0.42s

# Try the REPL
python3 apps/repl.py
> sub.bind('hello', 'world')
> sub.view('hello')
'world'

Now try the C99 version (the self-evolving meta substrate):

git clone https://github.com/SuperInstance/quilt-substrate-meta.git
cd quilt-substrate-meta
make && make test
# 36 tests, all passing

# Try the REPL
make repl
> bind greeting "hello, world"
> view greeting
hello, world
> journal
BIND greeting="hello, world" t=0

03Path 2: Cloudflare Worker

Deploy the LLM proxy. Get a free rate-limited endpoint. Use it from any of the web pages here (just point the base URL to your worker).

git clone https://github.com/SuperInstance/quilt-llm-worker.git
cd quilt-llm-worker
npm install
npx wrangler login
npx wrangler kv:namespace create "RATE_LIMIT"
# copy the namespace ID into wrangler.toml
npx wrangler secret put ZAI_API_KEY   # paste your key
npx wrangler secret put DEEPSEEK_API_KEY
npx wrangler deploy
# https://quilt-llm-worker.YOUR-SUBDOMAIN.workers.dev

Now your web pages can call your-worker/v1/chat/completions instead of the public one. Set window.QUILT_LLM_BASE = 'https://your-worker.workers.dev' in your browser console and refresh.

04Path 3: Full Stack

Clone the whole ecosystem. Deploy to your own domain. Vectorize your own canon. Save state in your own KV.

# 1. Clone the 24 repos
git clone https://github.com/SuperInstance/quilt-foundation.git
git clone https://github.com/SuperInstance/quilt-substrate-meta.git
git clone https://github.com/SuperInstance/quilt-ecosystem-web.git
# ... and 21 more

# 2. Deploy the web ecosystem to Cloudflare Pages
cd quilt-ecosystem-web
npx wrangler pages deploy . --project-name=quilt
# https://quilt.pages.dev

# 3. Set up the workers
npx wrangler kv:namespace create "STATE"
npx wrangler vectorize create quilt-canon --dimensions=768
npx wrangler r2 bucket create quilt-canon
npx wrangler d1 create quilt-gallery
npx wrangler d1 execute quilt-gallery --file=./schema.sql

# 4. Index the canon (200+ papers)
python3 scripts/index_canon.py --vectorize=quilt-canon
# 1046 chunks, ~5 minutes

# 5. Add your custom domain
npx wrangler pages domains add quilt.your-domain.com

Now you have a Quilt ecosystem on your own domain. The web pages, the workers, the KV, the Vectorize index, the R2 bucket, the D1 database — all yours. The cowboy's maxim extends: the watch is whoever is holding it.

05What you keep

All of it. The MIT license is at the root of every repo. The substrate is a contract; the implementations are interchangeable. The canon is yours to fork, extend, re-publish.

The collection is one idea, expressed in 24 doors. The doors are open. The cowboy's last ride is the ride you take.