Guides · 6 min read · Updated September 9, 2026
Automated directory submission with an AI agent (WebMCP)
How automated directory submission works with WebMCP: expose your launch package to an in-browser AI agent so it can read your product details and fill directory submission forms while you review.
An AI agent can already build your product. The reason it cannot launch it is not that the forms are hard; it is that the agent has nothing to fill them with. It does not know your tagline, it has to guess at a description and it has no way to get your logo. WebMCP fixes the data problem: it gives an agent running in your browser structured access to your launch package, so submitting to a directory becomes "read the package, fill the form, let the human check".
This guide explains what WebMCP is, what Launch to Win exposes through it, and how to use it to submit one or many products.
What WebMCP is
WebMCP is a proposal from the W3C Web Machine Learning community group. A web page registers tools with document.modelContext — named functions with a description and an input schema, the same shape as tools in the Model Context Protocol. An AI agent that runs in the browser (an extension, a built-in assistant, an agentic browser) can discover those tools and call them. The calls run in the page, with the user's session, so the agent can only do what the user could do.
It is early. The API surface has changed and will change again. Chrome and Edge ship it as an origin trial, which launchto.win has joined, so a current Chrome or Edge sees the tools without any flags. The WebMCP documentation keeps the browser table, the tool reference and a changelog current.
What Launch to Win exposes
When you are signed in on a plan that includes WebMCP (Maker or Shipaholic), every page on launchto.win registers three tools:
| Tool | Input | Returns |
|---|---|---|
list-packages | none | Your launch packages: name, slug, completeness. |
get-package | slug | Every field: name, tagline, short, medium and long descriptions, pricing, categories, keywords, alternatives, links, maker details, and URLs for the logo and screenshots. |
create-package | name and URL, plus any fields the agent already knows (tagline, descriptions, keywords, alternative to, launch date) | A new package. The URL triggers background enrichment; you finish it in the editor. |
Results come back as MCP tool results, so an agent that already speaks MCP needs no special handling. Every input field, result and error is listed in the tool reference.
Submitting a product with an agent
- Finish the package first. The agent fills forms from the package; an incomplete package produces an incomplete listing. Aim for a completeness score above 80%.
- Open launchto.win in the browser your agent runs in, signed in. The tools register on page load. Keep the tab open.
- Tell the agent what to do, in plain language: "Open uneed.best/submit-a-tool and fill the form using my Acme Analytics package from Launch to Win." A capable agent lists packages, gets the right one, reads the directory's form and fills each field from the matching package field — tagline into the tagline box, short description where the form wants one sentence, and so on.
- Review and submit. Check the category, the images and anything the form asked for that the package did not have. Then press submit yourself. Nearly every directory wants a human on that button and several have review steps that agents handle badly.
- Record it. Mark the submission done in your dashboard so the log stays accurate.
Submitting many products
This is where an agent earns its keep. The loop is the same per product, so the instruction is one sentence: "For each of my packages on Launch to Win, submit it to SaaSHub." The agent calls list-packages, then for each slug calls get-package and fills the form. Your job becomes reviewing a series of filled-in forms, which takes a minute each.
Go directory by directory rather than product by product; the agent stays on one form layout and you stay logged in to one site. The multi-product playbook explains why.
What it will not do
- It will not bypass a directory's rules. The agent acts in your browser with your session. Sites that require a badge, a newsletter signup or a manual review still require them.
- It does not submit server-side. WebMCP is about giving the agent your data in the browser. Launch to Win's own semi-automated one-click submissions are a separate Shipaholic feature.
- File uploads depend on the agent. The package exposes logo and screenshot URLs; whether the agent can attach them to a file input depends on what the agent can do in the browser. The Chrome extension handles uploads reliably if your agent cannot.
Privacy and control
The tools are only registered when you are signed in on an eligible plan. They read your own packages through the same session-authenticated endpoints the site uses. Nothing is exposed to other sites or other users, and the agent cannot act on your account from a page that is not launchto.win.
In short
Finish the package. Sign in on Maker or Shipaholic in the agent's browser. Tell the agent which package and which directory. Review, submit, record. Repeat per product with one sentence.
See the WebMCP documentation for browser support, how to verify the tools are registered and the full reference, and pricing for the plans that include it.
FAQ
Frequently asked questions
- Can an AI agent submit my product to directories?
- It can fill the forms. With WebMCP an agent running in your browser reads your launch package, opens a directory's submit page and fills each field from the matching package field. You review the result and press submit yourself; nearly every directory wants a human on that button.
- What is automated directory submission?
- Any setup where the typing is done for you. There are three levels: copy-ready fields, Chrome autofill that fills a form in one click, and an in-browser AI agent that reads the package through WebMCP and works through a whole list of products or directories from one instruction. None of them bypass a directory's rules or review.
- Which plans include WebMCP?
- Maker and Shipaholic. When you are signed in on one of those plans, every page on launchto.win registers list-packages, get-package and create-package with document.modelContext. The Free plan includes the Chrome extension but not WebMCP.
- Does the agent upload logos and screenshots?
- The package exposes their URLs; whether the agent can attach them to a file input depends on what it can do in the browser. If it cannot, the Chrome extension handles uploads reliably, and the two work together on the same form.
Do this once, not once per directory.
Build a launch package from your product URL, then submit it to 45+ directories with copy-ready fields, Chrome autofill or your AI agent.
Free plan available. No credit card required.