Forms for AI agents

AI Form Generator for Agents: Let Claude Build, Publish and Score Your Forms (2026)

AI form generator for agents: Claude builds, publishes and scores your forms with dForms. See how it works and what to check first.

Josue HernandezAI Specialist

AI Form Generator for Agents: Let Claude Build, Publish and Score Your Forms (2026)

An AI form generator should save you the whole job, not just the first draft. Most tools still stop halfway. You type a prompt, you get a list of questions, and then you click through an editor to fix logic, publish, and read the answers yourself.

That gap matters if you already run an AI agent. Your agent can write the questions in seconds. However, it can't finish the work if the form tool only talks to humans. So the form waits on you, and the leads wait on the form.

This guide shows a different setup. You connect an agent like Claude to a form builder made for agents, ask for the form in one message, and let the agent build it, publish it and read the scored answers. You stay in charge of what goes live. The agent does the clicking.

What an AI form generator does, and where most stop

An AI form generator turns a plain description into a working form. You describe the goal, for example "a lead form for a B2B agency", and the tool drafts the questions, field types and order.

That first draft is useful. Still, a draft is not a form people can answer. Somebody has to add conditional logic, connect the calendar, set up a webhook and publish. Then somebody has to read every answer and decide who is worth a call.

The three jobs behind every form

Every form you ship has three jobs:

  1. Build: questions, field types, logic and branding.
  2. Publish: a live link people can open and answer.
  3. Act on answers: score each response and send the good ones somewhere useful.

A typical AI form builder helps with the first job and leaves the other two to you. As a result, you save ten minutes of typing and keep the hour of setup and review.

What changes when an agent holds the keys

If an agent can reach all three jobs, the work changes shape. You ask once. Then the agent builds the draft, you check it, and it publishes. Later, the agent reads the responses, sorts them by score and tells you who to call first.

In other words, the AI is no longer a writing aid inside a form tool. Instead, the agent is the user of the form tool.

Why an AI form builder for agents works differently

An AI form builder for agents is built around one idea: the agent does the work, and you just ask. That sounds small. In practice, it changes what the product needs.

The agent needs an API, not a pretty editor

A human needs buttons and drag and drop. An agent needs clear routes, a key and predictable answers. So dForms exposes the same actions a person takes in the editor through an API. There are 44 routes, an API key per person, and a skill your agent can download to learn how to use them.

If you work in Claude or another MCP client, you can skip the API calls too. The dForms MCP at mcp.dforms.ai lets Claude create, publish, update and revert forms, list responses and read analytics. We checked it live with the Claude connector on October 21, 2026.

The agent needs guardrails that fail closed

Handing a key to an agent only works if the key can't do too much. Therefore, dForms keys follow an allowlist that fails closed:

  • A key can create, edit, publish, revert, duplicate and delete forms. It can also manage folders, destinations, notifications and the brand kit, read and delete responses, export CSV and read analytics.
  • A key can't manage keys, members, invitations or workspaces. It also can't change your profile or language, or connect and disconnect integrations.
  • New routes stay closed to keys until they are added to the allowlist.

Each key acts as the person who created it, with that person's current role. If you revoke the key, or that person leaves the workspace, the next call gets a 401.

The agent needs a way back

Agents make mistakes, and so do people. For that reason, edits live in a draft until someone publishes them. After that, a published form can be reverted. One honest note: dForms does not force a person to approve each publish. An agent with a key can publish. If you want a human check, tell your agent to leave the form in draft and send you the link.

How your agent builds, publishes and scores a form in dForms

Here is the full loop, from an empty workspace to scored answers. It takes four steps.

Step 1: Connect your agent

You have two options:

  • API key. Create one on the Developers page (/admin/developers). Your agent sends it as Authorization: Bearer dforms_.... You see the full key once, so copy it then. dForms only stores a hash of it.
  • MCP in Claude. Add the dForms connector at mcp.dforms.ai. After that, Claude can call dForms directly.

Owners and admins see every active key in the workspace and who created it. Each person can hold up to 10 active keys per workspace.

Step 2: Ask for the form in one message

Be specific about the goal, the questions and what counts as a good answer. For example, you can copy this prompt and edit the details:

Build a lead qualification form in dForms for my B2B agency.
Ask for: name, work email, company size, monthly budget,
main goal, and timeline to start.
Give each answer option points: higher budget and a start date
in the next 30 days score highest.
If the score is high, show my calendar booking step at the end.
Use my brand kit. Leave it as a draft and send me the preview link.

The agent builds the questions, the options with points, the logic and the ending. It also applies your brand kit, so the form looks like yours.

Step 3: Review the draft, then publish

Open the preview link and answer the form yourself once. Check the order, the wording and the logic. If something is off, tell the agent what to change, in plain words. When it looks right, say "publish it". Every form also records how it was born, api or web, so you can always tell which ones your agent made.

Step 4: Read scored answers and route hot leads

Once answers come in, dForms adds up the points on each one. High scores can go straight to your calendar booking step. In addition, you can sync answers to HubSpot or send them anywhere with a webhook.

Then ask your agent for a summary:

List this week's responses to the agency lead form,
sorted by score. For the top five, give me one line each
on why they look like a fit.

Want the scoring logic behind a good lead form? Read our guide on how to qualify a sales lead. And if an AI SDR handles your inbound, see how to build an AI SDR lead qualification form.

Form builder API: what to check before you hand it to an agent

Not every form builder API is ready for an agent. Before you connect one, check these points. They apply to any tool, dForms included.

Who the key acts as

A key should act as a real person, with that person's role. That way, an agent can never do more than its owner can. Also check what happens when someone leaves the team. In dForms, their keys stop working on the next call.

What the key can't touch

Look for a list of what a key cannot do, not only what it can. The risky actions are keys, members, workspaces and integrations. A safe API keeps those away from agents by default.

Limits and visibility

Agents can loop. So you want a rate limit that fails clearly, not silently. dForms allows 120 calls per minute per key. Past that, it answers 429 with a Retry-After header, so a well-built agent waits and tries again. dForms also tracks when each key was last used, so stale keys are easy to spot and revoke.

Undo

Finally, check for drafts and reverts. If an agent can only overwrite, one bad prompt can break a live form. With a draft and a revert, a mistake costs you a minute.

How AI form tools compare on agent access

Most big form tools now let an agent in through an MCP server, so agent access alone is no longer rare. The differences sit in pricing, limits and what happens after the answer arrives. We checked each competitor's own pages on October 22, 2026.

Tool Agent access Responses on the free plan
dForms API with 44 routes and the dForms MCP Not counted or billed
Typeform MCP server to create and edit forms and query responses "A limited number of responses per month" (pricing)
Tally MCP server, free on all plans Unlimited submissions within fair use (pricing)
Jotform MCP server for forms and submissions 100 submissions per month, 5 active forms (pricing)

So what sets dForms apart? It's built around the agent as the main user. Scoring per answer, routing to a calendar, HubSpot sync and webhooks all sit in the same place your agent already works. On top of that, dForms is free and open source under the MIT license, and you can self-host it.

Curious how agents use forms beyond lead capture? See how Claude Duet works with lead forms and how a Notion MCP gives agents memory through forms.

Frequently asked questions

What is an AI form generator?

An AI form generator is a tool that builds a form from a plain-language description. You describe the goal, and it drafts the questions, field types and logic. With a tool built for agents, like dForms, your agent can also publish the form and read the answers.

Can Claude build and publish a form for me?

Yes, Claude can build and publish a form in dForms. Connect the dForms MCP at mcp.dforms.ai, or give Claude an API key, and ask for the form in one message. Claude can create, update, publish and revert forms, and list the responses.

Is it safe to give my agent an API key?

A dForms key can only build and run forms, so it can't manage keys, members, workspaces or integrations. It acts as you, with your current role, and stops working if you revoke it. Keep in mind that a key can publish, so ask your agent to leave forms in draft if you want to review first.

Can I make sure a person reviews the form before it goes live?

Yes, as a practice you set with your agent. dForms keeps edits in a draft until someone publishes, and you can revert a published form. However, dForms doesn't enforce an approval step, so tell the agent to stop at the draft and send you the preview.

How does the form score each answer?

You give each answer option points, and dForms adds them up for every response. Then you can send high scores to your calendar booking step, sync answers to HubSpot, or post them to a webhook.

How much does dForms cost?

dForms is free, and responses are not counted or billed. It's also open source under the MIT license, so you can self-host it if you prefer.

Let your agent finish the job

A good AI form generator shouldn't hand you a draft and walk away. Instead, it should let your agent build the form, publish it and tell you which answers deserve a call. Connect your agent, ask for your first form in one message, and keep your time for the people who actually fit.

Keep reading