Cube Forms

Product Introduction

1. Definition: Cube Forms is an agent-native form platform, a technical category of API-first, AI-accessible web forms. It provides a single URL endpoint that serves both human-readable web pages and machine-readable specifications (Markdown/JSON Schema) for AI agents.
2. Core Value Proposition: It exists to bridge the gap between human and AI-driven form interactions, enabling AI agents (like Claude, ChatGPT, or custom assistants) to programmatically read, fill, and submit web forms on behalf of a user without requiring a browser or manual intervention, while maintaining a seamless human-in-the-loop experience.

Main Features

1. Dual-Purpose Form Links: Every Cube form has one URL that serves two distinct content types based on the request. For humans, it renders a standard HTML form. For agents, appending .md to the URL or sending an Accept: text/markdown header returns a complete Markdown specification of the form, including all questions, accepted input types, and submission instructions. A .json suffix provides a structured JSON Schema.
2. Agent Submission via HTTP/MCP: AI agents can submit form data directly via a simple POST request with a JSON payload to the form's URL. For integrated agent environments, it supports the Model Context Protocol (MCP), offering tools like submit_form and prefill_link. File uploads are handled via multipart form-data, base64 encoding, or public HTTPS URLs.
3. Owner Management Server: Form creators can connect their AI agent to an "owner" MCP server (https://ui.usecube.io/mcp/owner) via OAuth 2.1. This grants the agent a suite of tools for full lifecycle management, including create_form, update_form, delete_form, list_answers, export_answers, and configuring integrations like email, webhooks, and Slack, all without manual browser use after initial authentication.

Problems Solved

1. Pain Point: Eliminates the friction and inefficiency of AI agents being unable to interact with traditional web forms designed solely for human browsers, which often involve CAPTCHAs, complex JavaScript, and manual copy-pasting.
2. Target Audience: Developers building AI agents and automation workflows; power users of AI coding assistants (Claude Code, Cursor, ChatGPT) who want their agents to handle administrative tasks; product managers and support teams seeking to automate intake processes like bug reports, sales inquiries, or waitlist sign-ups.
3. Use Cases: An AI assistant automatically filling out a client intake form with known user details; a coding agent submitting a bug report directly from an error log; an HR agent managing job application responses and exporting data; creating and publishing a feedback form entirely through a chat interface.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):

   • Pros: Unifies human and agent interaction into a single, simple link. Removes dependency on browser automation (e.g., Puppeteer, Playwright). Offers robust programmatic access via both plain HTTP and the standardized MCP. The free tier lowers the barrier to entry. Human oversight is preserved through prefilled links and receipt review.
   • Cons: Requires the form creator to use Cube's platform; cannot be applied to existing third-party forms. The agent's ability to act autonomously on behalf of a user depends on the user first authenticating the agent via OAuth for owner features. Relies on agents correctly interpreting the Markdown/JSON Schema instructions.

2. Key Alternatives & Differentiation:

   • Traditional Form Builders (Typeform, Google Forms): These lack native, standardized APIs for AI agents. Automation requires unofficial scraping or complex API workarounds. Cube is built from the ground up for agentic interaction.
   • Browser Automation Tools (Playwright, Selenium): These simulate a human user in a browser, which is resource-intensive, brittle against UI changes, and often blocked by CAPTCHAs. Cube provides a direct, lightweight API, avoiding the "browser tax."
   • Generic HTTP API Platforms: While agents can POST to any API, Cube provides the crucial structure—the self-documenting form schema (Markdown/JSON)—dynamically from the same URL, eliminating the need for separate, static API documentation.

Frequently Asked Questions (FAQ)

1. How do AI agents fill out Cube Forms? AI agents fill Cube Forms by first fetching the form's Markdown specification (by adding .md to the URL), then submitting a JSON payload via an HTTP POST request or using the submit_form tool via the Model Context Protocol (MCP).
2. Can a human still fill out a form made for an AI agent? Yes, every Cube Form has a single link. When opened in a standard web browser, it displays a normal, user-friendly web form that anyone can fill out manually, ensuring full human accessibility.
3. Is Cube Forms free to use? Yes, according to the provided documentation, Cube Forms is currently a free product for creating forms and having agents fill them in.
4. How do I connect my AI assistant to manage my forms? You connect by adding the Cube owner MCP server URL (https://ui.usecube.io/mcp/owner) as a custom connector in your AI platform (like Claude, ChatGPT, or Cursor) and then completing a one-time OAuth authentication flow in your browser.
5. What happens to the data submitted through Cube Forms? Form owners can receive submissions via the Cube dashboard, exported files, or configure automated integrations to send data directly to email, a custom webhook endpoint, or a Slack channel, giving them full control over their response data pipeline.