Our website, ready for an AI agent to use.

An AI agent can fill in the contact form with what the user asked for. A person always sends it.

The problem

“I asked my assistant to write to that company and it did not know how.” “It gave me the website address and I ended up filling in the form myself.” “I do not want an AI sending anything in my name without me seeing it.”

It happens to any company whose website is built only for people with a browser: whoever arrives through an AI agent is left halfway.

How it is done today

The visitor fills in the contact form by hand. If they arrived while talking to an AI agent, they have to leave the conversation, find the form and type again what they had already explained.

How we solve it

  1. The user asks their agent in one sentence, for example “ask Julditec for an AI Diagnostic for my company”.
  2. The agent drafts the message from what the user asked for.
  3. It fills in the website's contact form with that message and the person's details.
  4. The form is left filled in and in view. The agent does not send it.

Result

It already works on this website: an AI agent can fill in our contact form, and a person always sends it.

What it takes

  • Your website and its form

    Access to the website's code and to the form you want to open to agents. It is added on top of what you already have.

  • Your content, in order

    What you offer and where it is explained. An agent recommends better what it can read without guessing.

  • What an agent may do, in writing

    Which actions are opened to it and which are not. On ours, filling in is allowed; sending is not.

  • A person who sends

    The last step stays with whoever signs the message. That way nobody receives or sends anything a person has not seen.

How it is built

The website registers its tools in the browser with WebMCP (navigator.modelContext). One of them, prepare_contact_message, fills in the contact form with name, email and message. It does not send it: the person reviews it, passes the reCAPTCHA and presses “Send”.

For an agent that is not in the browser there is a link that opens the form already filled in. The data goes after the hash in the URL, so it never leaves the person's browser. The instructions for agents are published in llms.txt and in a skill under /.well-known/agent-skills/.

There is also an anonymous MCP server at /mcp for querying the website's content. That server does not have the contact tool: without a browser there is no person to review and send.

Frequently asked questions

Can an AI agent send the form for me?

No. The agent fills it in with what you asked for. A person always sends it.

Why is the agent not allowed to send it?

Because the message goes out in your name. If you review it and send it yourself, nobody sends anything a person has not seen.

What do I ask my agent?

What you need, in your own words. For example: “ask Julditec for an AI Diagnostic for my company”. The agent fills in the form and you send it.

Related automations

  • Interface design

    The agent builds, reviews and polishes interfaces with our own design guide. A person chooses the direction.

  • Development from an issue

    Assign the issue to the agent and development starts on its own. A person reviews and merges.

Part of AI agent development

The next step

Tell us how you work today and we will tell you where to start.