Run OctaFuse on Cloudflare without managing a server

Independent developers and small teams often combine coding plans, token plans, official APIs, aggregators, and self-hosted models. Every upstream adds another Base URL, key, quota, and price to maintain.

A self-hosted gateway turns those scattered resources into one client-facing URL and user key. OctaFuse can run on Cloudflare Workers + D1, so small deployments do not need a continuously managed server.

Deployment shape

AI Client
   │ user API key
   ▼
Proxy Worker ───┐
                ├── shared D1 database
Admin Worker ───┘
   ▲
   │ browser login / MASTER_KEY
Operator

Proxy serves model and Tools traffic. Admin provides the configuration and operations control plane. Both Workers must bind to the same D1 database.

From an empty database to the first request

The main path is:

1. Prepare a Cloudflare account, Git, Node.js, and npm.
2. Clone the repository and authenticate Wrangler.
3. Run the bootstrap flow to create D1, Proxy, and Admin.
4. Sign in to Admin and rotate the development Master Key immediately.
5. Configure a Provider, Model, and Route.
6. Create a user and issue a user API key.
7. Verify /v1/models or send a Chat Completions request.

Use the current commands

Deployment scripts and migrations evolve with each release. Use the latest quick start and Cloudflare deployment guide for executable commands.

An empty /catalog/models response immediately after deployment is expected: the database does not yet contain your Providers or active Routes.

Keep ADMIN_PASSWORD, the Admin API MASTER_KEY, and user API keys separate. They authenticate different surfaces and should never be substituted for one another.

• Current quick start
• Cloudflare deployment guide

Learn more about OctaFuse or explore the source:

• OctaFuse website
• GitHub repository

If OctaFuse is useful to your project, consider giving the repository a Star. Your interest and feedback help us keep improving deployment and the self-hosted experience.