MCP server for AI agents

What it looks like once it's connected

Say you type "where should myapp.com go next?". The agent finds that project, asks for sites it hasn't touched, highest DR first, and can drop the paid ones or the ones that need an account. Then it shows you the list, with DR, pricing and what kind of form each site has.

You pick one and ask "how do I submit to aitoolsdirectory?". It reads you the submit URL, the steps and the tips other people left, the same stuff that's on that site's page here. You fill in the form, say "mark it submitted", and it writes that to your account with a note on what confirmation you saw.

Two weeks later you ask "which ones are still in review?" and it pulls them up. These marks are the same ones you see in your tracker on the site, so both sides always match.

Who it's for

Anyone already using an AI agent. If it writes your code or does your research, it can do your launch work too. You stop switching between the website and the terminal, and that's the whole point.

Connect in three steps

1. 1.Sign in, open Account and find MCP for AI agents.
2. 2.Give the token a name, like "Claude Code on MacBook", and click Generate token. Copy it right away. It's shown once and starts with wts_.
3. 3.Add the server to your agent. For Claude Code, run this in your terminal with your token in place of wts_....

claude mcp add --transport http wheretosubmit https://wheretosubmit.org/api/mcp --header "Authorization: Bearer wts_..."

For Cursor and other clients, drop this into the MCP config file. Cursor reads ~/.cursor/mcp.json. Claude Code also reads .mcp.json in your project root.

{
  "mcpServers": {
    "wheretosubmit": {
      "type": "http",
      "url": "https://wheretosubmit.org/api/mcp",
      "headers": { "Authorization": "Bearer wts_..." }
    }
  }
}

The Account page shows the same snippet with your token already in it, ready to copy. Keep the token out of git and treat it like a password.

What the seven tools do

The names are for the agent. You don't need to remember them, just know what it can do.

list_projects

Lists your projects with each product's domain, and tells the agent your plan. Usually the first call it makes.

project_status

The sites you've already marked for a project, newest first. Filter by status, or show only the ones with something left to do, like a badge you haven't added.

next_targets

Sites the project hasn't marked yet, highest DR first. Filter by price, minimum DR, category, form type or keyword.

get_site

Everything about one site: submit URL, how to submit, tips. Pass a project and you also get your mark for that site.

mark_status

Records results for up to 25 sites at once, each with a note on what the agent saw, like "page said Thanks, we'll review".

create_project

Makes a new project. It checks your existing ones first and asks if something looks like the same product, so you don't end up with duplicates.

update_project

Renames a project or fills in its product URL.

How far you get on the free plan

Making a token and reading data works on every plan. The sites, DR and pricing you can see are the same as when you're signed in on the website.

On the free plan, Pro-tier sites come back without the submit URL, method and tips, same as the website. Writing marks to your account, and creating or editing projects, need Pro.

You can have up to 5 tokens, and browser extension devices count. Remove one on the Account page and it stops working right away.

Security

We store a hash of each token, never the token itself. If you lose it, make a new one and delete the old one.

A token only works for your own account, and only on the MCP endpoint. Ask for someone else's project id and you get "project not found", the exact same answer as for a project that doesn't exist. Nothing to guess from.

There's no call limit, so an agent can run all afternoon without getting cut off.

Things you can just say

• "List my projects."
• "What should I submit myapp.com to next? Free ones only."
• "Mark aitoolsdirectory.com as submitted, I saw the confirmation page."
• "Which sites are still waiting for my badge?"

The agent picks the tool on its own. Plain words are enough.

Want it to fill the forms too?

This server only handles "where to submit" and "what's done". It doesn't touch forms. To have the agent actually fill them in, site by site in your browser, add the agent skill. The skill walks it through the work, and the results come back through this server just the same.

Every parameter and error code is in the docs on GitHub.

• Agent skill for submissions
• MCP docs on GitHub