A researched list of launch sites and directories, a plan that starts with the strongest free ones, and a prompt for your agent to fill in the forms while you approve. With a zero budget, it handles about 22 on its own and tells you what to do on the rest.
Launch your Startup is a free kit for anyone who built an app, website or SaaS and wants it to show up on the internet. It brings together 59 launch sites and directories (like Product Hunt, G2 and AI directories) that send visits and links back to your site, the so-called backlinks, which help with Google. A script builds the plan in the right order and keeps track of progress, and a ready-made prompt has your AI agent fill in the forms for you. What you need: Python 3, Claude Code with the Chrome extension, and an email address just for sign-ups.
The list says where to go, the script says in what order, the agent does the boring work. You only show up where you can't be replaced.
Each site with its submission link, estimated strength, cost (free or paid), whether it requires a badge on your site and whether it needs you. Includes 3 sites for the Brazilian audience.
lancar.py reads your brand sheet, splits the sites into 4 groups and creates a spreadsheet (CSV) with the status of each one.
It interviews you, builds the sheet with real numbers only, shows you the plan and fills in site by site. It stops at anything that costs money or touches your site.
Two approval gates before any submission. Nothing goes out without your yes.
Free, no badge and no need for you. The agent fills them in on its own, from strongest to weakest.
Asks for a badge on your site or is paid (within your budget). The agent shows the cost and waits.
Launches with a set date and communities (Product Hunt, Hacker News and the like). The agent gets the text ready; you post it.
Paid only and you have no budget, or it asks for a badge and you said no.
Sorted by estimated site strength (high, medium, low). Strength is an estimate based on the site's size and reputation, not an official number: a link from a strong site counts more for Google. 46 of the 59 were opened and checked on Nov 7, 2026; 13 blocked automated reading and came from public sources (the conferido_em field in sites/sites.json says which). Prices and rules change: the agent checks at submission time.
On a phone, swipe the table sideways to see cost, badge and who does it. A badge is a small emblem with a link to the directory that some sites ask you to put on your site in exchange for the listing. The 3 Brazilian sites need you (they are communities or company registries).
| # | Site | Strength | Cost | Badge on your site | Needs you | Type |
|---|
Nei walks through the list, the four-group plan and the agent filling in the sign-ups (in Portuguese).
Nothing paid. Just tools you probably already have.
For the plan script. Works on Linux, Mac and Windows.
git clone https://github.com/inematds/lance-sua-startup cd lance-sua-startup python3 lancar.py lista
The agent is Claude Code (it requires a paid Claude plan). It controls your browser through the Claude in Chrome extension: install it from the Chrome Web Store, log in and, inside Claude Code, run /chrome to connect.
# install (Linux/Mac) curl -fsSL https://claude.ai/install.sh | bash # open in the kit folder cd lance-sua-startup && claude
Create an inbox like launch@yourbrand.com and keep it open in a tab. That's where the confirmation links arrive.
512×512 logo and 1 to 3 screenshots
inside the marca/ folder
(e.g. marca/logo-512.png and
marca/captura-1.png)The commands run in the kit folder. The example uses the ready-made sheet in exemplo/, so you can see the result before filling in your own.
List only the sites the agent can do alone, or only the free ones, or only the Brazilian ones.
python3 lancar.py lista --agente --gratis python3 lancar.py lista --mercado br
The example sheet belongs to a real INEMA project. The script checks the sheet and builds the plan in a test folder.
python3 lancar.py checa --marca exemplo/marca.json python3 lancar.py plano --marca exemplo/marca.json --saida exemplo/trabalho # opens exemplo/trabalho/plano.md and exemplo/trabalho/rastreador.csv
Open Claude Code in the kit folder (run claude in the terminal) and paste the part between the --- lines of agente/PROMPT-LANCAMENTO.en.md. It creates marca/marca.json, interviews you one question at a time and runs checa until it says ficha OK.
# at the end of the interview, you type:
brand sheet approved
The agent runs the plan and shows how many sites fell into each group. With a budget of 0 and no badges (as in the example sheet), today's result is: 22 the agent does alone, 23 are your turn and 14 stay out.
python3 lancar.py plano # the agent runs it; uses marca/marca.json plan approved # you type
For each site in group A: it opens the link, creates the account with the sign-up email, fills in the form from the sheet, confirms the email and logs it in the tracker. You don't need to type these commands: the agent runs them, after the real plan (trabalho/).
# the agent runs, after each site: python3 lancar.py marca pitchwall enviado --nota "free listing" python3 lancar.py status
status lists the sites marked as your turn (sua-vez) and the draft the agent left for each one in trabalho/rascunhos/. In 1 to 2 weeks, confere opens each listing and looks for your domain; if it finds it, it marks it as live (no-ar).
python3 lancar.py status python3 lancar.py confere
It's written in the prompt and in the plan. If any of the items below shows up, the agent marks the site as your turn and moves on to the next one.
No charge, plan, "boost" or paid queue, not even US$ 1, without your yes for that site.
Badges or backlinks only go in with your yes, site by site.
Only texts and numbers from the sheet. A number without a source is rejected by checa.
Captcha, phone, ID document or video: it stops and calls you. And it never says it submitted what it couldn't submit.
No posting in communities with your personal account. Product Hunt and Hacker News are yours.
Before signing up, it checks whether the product is already on the site. The tracker keeps what was done and the proof of each submission.