No rewrites. One script tag.

The human interface
for your product.

AgentHint turns your product into a conversational interface. Language is the most powerful UI — AgentHint finally makes it practical. Instead of more buttons and flows, let your users just ask.

Smart
Understands intent.
Fast
Speeds up real tasks.
Trusted
Calls only what you allow.
Drop-in
Works on any site.
How it works

From repo to chat widget in three steps.

AgentHint reads your frontend the way you would — by following the code — and wraps your existing API calls as callable tools for the assistant. Your backend, your auth, your domain.

Step 01

Connect your repo

Point AgentHint at your frontend GitHub repo. We clone it read-only and use AI to map API calls, env vars, and how your app authenticates.

Step 02

Review your SDK

Every detected call becomes a typed SDK method. Toggle what the assistant can do and override any env vars that differ in production.

Step 03

Embed the widget

Paste one script tag before </body>. The widget loads on demand on your own domain, using your user's existing session and cookies.

Why AgentHint

Not a chatbot. A better way to interact.

No rewrites

AgentHint wraps what's already there. You don't need a new backend, a new auth model, or a new front-end framework.

Your auth, your domain

The widget runs on your site and uses the user's existing session. Calls go to your API with the same credentials a logged-in user already has.

Deterministic actions

The assistant can't invent endpoints. It calls only the SDK methods you enabled — reviewed, typed, and traceable.

Built for humans

Users don't need to learn your menus. They ask. AgentHint finds the right tool, asks for any missing info, and gets out of the way.

Give your product a voice.

Set it up in an afternoon. Live on your domain by tomorrow.