agentic-ui-vueAI actions for your Vue app, on your terms

Let an AI assistant help people use your Vue app — while your app stays in control.

Keeps up with the page

Capabilities appear when a component mounts and disappear when it unmounts, so the available actions always match what is on screen.

Checks before acting

Every request is validated and permission-checked. For mutating or destructive actions, the user approves the exact payload — a digest remembers what was already approved.

Chooses the right item

If the page has two copies of the same widget, each can register a target. The AI receives a _target enum so it can send the action to the right one.

Handles larger apps

With 50+ capabilities, provide 3 helper meta-tools (search / describe / execute) and let the AI find other capabilities only when it needs them.

What it is

agentic-ui-vue lets you decide exactly what an LLM (the AI behind an assistant) may do in your Vue app. Each component can declare capabilities: actions with defined inputs, clear descriptions, and risk labels, such as listing todos, adding an item, or saving a form. Those risk labels let users approve changes before they happen.

The AI does not need to inspect the page or imitate mouse clicks. It asks your app to run one of the actions you have made available. Before anything runs, the request goes through the same safety path — validate → permission → confirmation → execute — and the result goes back to the AI so it can respond or continue.

// A component says what the agent is allowed to do here:
useAgentCapability({
  name: 'todos.add',
  description: 'Add a new item to the todo list.',
  inputSchema: { type: 'object', properties: { text: { type: 'string' } }, required: ['text'] },
  risk: 'mutating',
  execute: ({ text }) => todos.push({ text, done: false }),
})

This is not browser automation: the AI does not click buttons or read the page layout. It calls the same functions your interface already uses.