Tether
DocsGitHub

Start here

Which door?

There are two ways into Tether, and the honest question that separates them is whether you already have an AI you are happy with.

You already have one

You run an agent. It has tools, a model you chose, a provider key you pay, and a UI you built. What you do not have is anything stopping it from doing something expensive at three in the morning.

Tether goes in front of your tools. Reads carry tenant scoping and an audit trail. Writes stop and wait for a person. Your agent, model and UI stay exactly as they are, and tool results never pass through Tether to a model provider.

You do not

You have a product, data and a brand. You have never written a prompt and do not intend to. You want the thing your customers talk to, not a model to operate.

Tether brings the model, the billing and the surface. Five steps, each useful on its own, and nothing is live on your site until the last one.

You build software other businesses run

You are a software provider: a booking platform, a shop platform, any product with many business customers who each have customers of their own. You connect your API once, and every business on your platform gets an agent its customers can ask, in Georgian, to do something, with each business's own staff approving where its policy says so. It is the second door, taken once for all of your businesses.

What both doors share

They are not two products. They read and write the same six objects, so a first-door customer who later wants the surface, or a second-door customer who later hires an AI team, keeps everything they have already set up.

Bring your AIUse Tether AI
Model and provider keyyoursTether's
Your agent and UIunchangedTether's widget, or your own
Tool results reach a model providernoyes, Tether's
Writes stage and wait for a personyesyes
Audit trail, budgets, tenant isolationyesyes

Both are hosted. Neither is a different codebase.