What you own

Typillar is built around a deliberate ownership split. Understanding it is the fastest way to understand the whole product, because it explains where things run, who can see what, and what happens if you leave.

You own

• Your Cloudflare account — the compute and hosting your product runs on.
• Your GitHub repository — the source of truth for all generated code.
• Your databases & hosting — D1 databases, Durable Objects, and the deployed Workers all live in your account.
• Your models & API keys — the provider that generates code is yours.
• Every token of inference — billed to you, run on your resources, never proxied through ours.

Typillar owns

• The great code — the agent system, the planning logic, the orchestration that turns intent into shipped software.
• The PM + agent control plane — the always-on product manager and the builder agents that drive the loop.

Why split it this way

Two reasons.

No lock-in. Because the product, its source, its data, and its accounts are all yours, leaving Typillar is a non-event: the running app keeps running, the repo is still in your GitHub, the data is still in your Cloudflare. There’s nothing to export and no hostage to negotiate.

A clean trust boundary. Typillar never needs to host your data or run your inference, so it doesn’t. The control plane coordinates work; your infrastructure executes and stores it. That boundary is what makes the security model simple to reason about — see Permissions & access.

What this means in practice

Question	Answer
Where does my app run?	Your Cloudflare account.
Where does my code live?	Your GitHub repository.
Where is my data stored?	Your Cloudflare databases.
Who pays for inference?	You, on your model provider or Workers AI.
What if I stop using Typillar?	Everything keeps running; it’s all yours.

Related

• The build loop — where each stage runs.
• Zero inference on our servers — the inference half of the split.