The Launch Kit

Your coding agent submits your product to startup directories, in your own Chrome, signed in as you. The kit tells it which ones are really free, plans the run and keeps the record.

Not on sale yet. It is built and we use it on LaunchGrill itself. Leave an address and you will hear once, when it opens.

89free routes

79free on a condition

51paid only or gone

301not read yet

520 playbooks. A route counts as free only after its submit and pricing pages were read page by page, and each carries the date. The ones not read yet are in the catalogue and a run never uses them.

Why not just tell your agent to submit you everywhere?

You can. It fills forms well. It does not know which forms are worth filling.

Your agent aloneYour agent with the kit

Searches for “startup directories” and gets lists full of dead sites.Starts from 168 routes that were read page by page, each with its date.

Meets the fee, the badge or the link-back rule halfway through the form.Reads the fee, the badge rule and the queue before it opens the form.

Works each form out again on every run.Follows the playbook: the fields, their limits and the traps someone already hit.

Forgets what it submitted when the session ends.Records every outcome, and never submits to the same directory twice.

Writes a new pitch each time and invents claims along the way.Writes from one brief and claims only what you listed in it.

How it works

01
Describe your product once.
A brief holds the facts, the copy and the claims you allow. If you came from a grill it arrives drafted from your page, with what the grill says to fix. You answer once what your agent may do: create accounts, confirm emails, submit without review.
02
Plan a run of 10.
The planner picks free routes that fit your categories and prints the prompt for your agent. If fewer than 10 fit, the target is smaller. It is never filled up with unread or paid directories.
03
Submit, record, resume.
Your agent works through the list in your Chrome and records each outcome before the next. It stops for captchas and anything that costs money. A later session picks up what is left.

What is in it

A playbook per directory. What the free tier gives, the condition if there is one, the queue, the form’s fields and the traps, with the date it was checked.
The link, as measured. Whether a listing’s link is followed is read from the listing’s HTML, not from the directory’s claim. 84 are measured so far; the rest say “unknown”.
A brief written from your grill. Taglines, descriptions and categories drafted from your page.
A planner and a tracker. A run is a target of new confirmed submissions. Drafts and listings that already existed do not count.
A local dashboard. What needs you, where every submission stands, the catalogue and the run’s prompt. It runs on your machine only.
Setup guides. The first session, the browser connection and the permissions, led by your agent.

What still needs you

Captchas and two-factor codes. The agent stops, says exactly what to do, and moves to the next directory.
Anything that costs money. It never pays. A directory with only a paid route is recorded as such, with the price it saw.
Badges and links back. Some directories are free only with their badge on your site. You decide once: ask each time, never, or allowed.
How you are described. A maker bio or “why I built this” is written only from wording you approved.
Product Hunt. It bans bots, so it is a checklist you do by hand.
Saying a listing is live. That needs the public URL and your word that you opened it.

What a record looks like

Nothing is ever just “done”. Each state is tied to what was seen, and only the first four count toward a run’s target.

In reviewThe directory said it received the submission; moderation is pending.
QueuedThe directory confirmed a place in a queue or waiting line.
ScheduledThe directory confirmed a launch date.
LiveThe public listing was opened, carries no pending label, and the founder has looked at it.
Needs youPrepared work waits on one exact action from the founder (captcha, 2FA, a login, a decision).
Paid onlyThe only usable route costs money; nothing was bought.
DeclinedThe founder said no to a condition (a badge, a link back).

We ran it on LaunchGrill first

70 directories so far: 6 live, 6 scheduled, 3 queued, 24 in review, 8 declined because they wanted a badge or a link back, 2 paid only. The report says what “free” turned out to mean for the first 20, and 225 directories have a public page with their dated facts.

A sample. The playbooks add the form’s fields, the steps and the traps.

What you need

Node.js 18 or newer, Chrome, and a coding agent that can read local files and drive a browser you can see. An agent without a working browser connection cannot do this. Check yours now, in any folder that holds a text file:

Read one text file in this folder and tell me its first line. Then use your browser tool to
open https://example.com in Chrome and tell me the page title and its main
heading. Do not sign in, fill a form or change any setting. If you cannot do
either action, tell me which capability is missing.

Both worked: the basics are there. Only the file: connect a browser to your agent first. Your agent’s own usage is billed by its provider.

Questions

What do I get?

A folder you open in your coding agent: 520 directory playbooks, the rules your agent works under, a planner, a tracker, a local dashboard and the setup guides. It runs on your machine. Nothing is uploaded to us.

Why not just tell my agent to submit me everywhere?

You can. It will fill forms well and guess at everything else: which directories still exist, which “free” ones want a badge or sell the link, and what it already submitted last week. The kit gives it dated facts per directory, a plan and a record.

Does my agent do the submissions, or do I?

Your agent, inside the permissions you give it once per product: whether it may create free accounts, confirm verification emails, and send the final submission without showing you the draft. With no permissions given it prepares everything and asks you each time. Some agents will not create accounts or sign in for you whatever you allow; then it hands that one step to you and carries on.

What does it never do?

It never pays, never ticks a paid option, and never solves a captcha or a two-factor prompt: it stops, tells you exactly what to do, and carries on with the next directory. It never automates Product Hunt, which bans bots; that one is a checklist you do by hand.

Does it run while I’m away?

No. Your agent works in your session, in a Chrome window you can see. Nothing runs in the background. A saved run picks up where it stopped.

Which agents work with it?

Any agent that can read local files and drive a browser you are signed in to. Claude Code, Codex and Cursor can once their browser connection is set up. The check on this page tells you whether yours is ready.

Will it bring traffic or customers?

We can’t promise that, and our own numbers say to expect little at first: the two live listings from our own run had sent no visitors when we last measured. What you get for certain is a dated record of what was submitted, what went live, and what each listing’s link really is.

How many products can I use it for?

As many as you own. Each product gets its own folder with its own brief, permissions and tracker.

What does it cost?

No price is set. It is not on sale yet; leave an address below and you get one email when it opens.

Why not just tell your agent to submit you everywhere?

How it works

Describe your product once.

Plan a run of 10.

Submit, record, resume.