LaunchRepoEvery “best startup directories” list you will find this year ranks platforms by a number nobody can check and treats a listing as a result. This article does neither. It explains how the 346 directories in the LaunchRepo catalog are grouped, which ones you will recognise, and how to decide — per product, in a few minutes each — which ones deserve your time. No traffic figures, no authority scores, no promise that a listing moves your rankings. Just a method you can run yourself, or hand to your AI coding agent.
What a directory is, and what it is not
A startup directory is a website that publishes a page about your product: name, tagline, description, screenshots, a link back to you, sometimes votes or comments. Product Hunt is the famous one; hundreds of smaller sites follow the same pattern with different audiences and different rules.
Three things are worth stating before choosing any of them:
- The directory decides what it publishes. You submit; a moderator, a queue or an algorithm decides whether and when the page goes live. On many free tiers that takes weeks, not days.
- A listing is not a backlink guarantee. Some directories link to your site with a normal, crawlable link. Others add
nofollow, route the link through a redirect, or render it with JavaScript only. You cannot tell from the outside without looking. - A listing is not traffic. Some platforms send a steady trickle of relevant visitors; most send almost none. Anyone who gives you a number per directory without measuring your product on that directory is guessing.
If you accept those three points, “the best directory” stops being a global ranking and becomes a question about fit: which platform’s audience wants your product, and which one will actually publish it for free.
The six categories in the catalog
The LaunchRepo catalog lists 346 directories, each with its own playbook. Every entry carries one of six categories. The counts below are from the catalog as of 17 September 2026.
| Category | Entries | What lives here |
|---|---|---|
| Product launch | 104 | Launch-day platforms with votes, comments or “new this week” feeds. The Product Hunt family. |
| AI tools | 100 | Directories that only take AI-related products. The fastest-growing category since 2023. |
| Software tools | 85 | General software catalogs and alternative-finders, no launch moment required. |
| B2B software | 32 | Review and vendor platforms for buyers; often need company facts, sometimes paid tiers. |
| Company profile | 18 | Startup databases and ecosystem maps: organisation entries rather than product pages. |
| Community | 7 | Places where a launch is a post in a community, with the rules of that community. |
The categories matter more than the total. An AI writing assistant can reasonably target four of them; a developer library for Kubernetes fits maybe two. The counts also explain why a “submit to 300 directories” offer is almost never right for a single product: a third of the catalog only takes AI tools, and the B2B review platforms want facts that a solo side project does not have.
Product launch platforms
These are the directories people mean when they say “launch”. Product Hunt, Uneed, Fazier, TinyLaunch, IndieHunt, Pitchwall, LaunchBoard, PromoteProject and SubmitHunt are all in this group, and so are newer ones like Smol Launch, Nick Launches, StartupTrusted, Startup Buffer, ComingUp and Launching Next. They share a rhythm: you pick or are assigned a launch date, the page goes live that day, and the platform’s ranking rewards early activity. The playbooks for Product Hunt, Uneed, Fazier, TinyLaunch and Pitchwall are public on this site if you want to see what a preparation checklist looks like.
Two practical notes. First, several of these platforms let you schedule rather than publish immediately, and a scheduled date is not a live listing — your tracker should keep the two apart. Second, many launch platforms want a badge on your website in return for a free slot; decide once per product whether you are willing to add one.
AI tool directories
If your product uses a model in any meaningful way, this category alone gives you a hundred candidates, among them ToolPilot and Stork.AI from the catalog. They tend to be easier to get into than launch platforms and more demanding about screenshots and a clear “what does the AI actually do” sentence. Most will reject a generic SaaS with an AI feature bolted on, so read the acceptance criteria in the playbook before submitting. Our page for AI tool founders goes deeper on the fit question.
Software tools and alternative-finders
SaaSHub, AlternativeTo, ToolDirs, SaaS Browser and AppRater belong here. There is no launch day; you create an entry that sits next to the products people already compare you with. The work is mostly accurate categorisation and naming your alternatives. The SaaSHub playbook shows what “prepare” means on such a platform.
B2B software and reviews
GoodFirms, SaaSworthy, OMR Reviews, F6S and SourceForge live in this group. They ask for company details, sometimes want a verified business email, and often have a paid tier that is presented as the default. The free route usually exists but is less prominent. If you are launching as an individual rather than a company, expect a few of these to be a poor fit — and record that as “not a fit” rather than forcing it.
Company profiles and communities
StartupBlink is a typical company-profile entry: an organisation record in an ecosystem map. Peerlist’s launchpad is the catalog’s example of a community launch. Both are small groups; neither should be your first stop, but they are cheap to add once the product page copy exists.
How to choose without wasting a week
The naive approach is to work down a list alphabetically. The expensive one is to buy “200 submissions” and hope. The method in the catalog’s PRIORITY.md sits in between and fits in roughly five minutes per directory. In my own words:
- Check fit and the free route first. Does the directory take your category at all? Is there a free way in, or only a paid one? If you have already submitted there — even as a draft — that entry goes to follow-up, not to a new queue.
- Look at three public listings in your category. Ideally one established product and two recent ones. Note the exact URLs and the date.
- Search Google for those exact pages. Use
site:and the product name plus the directory name. A hit is evidence that the directory’s pages do get indexed. A miss is unknown — it is not proof that nothing is indexed, and a result from another search engine does not count as Google evidence. - Inspect the listing page itself. HTTP status, redirects,
X-Robots-Tag,robots.txt, meta robots and the canonical tag, in the raw HTML and in the rendered DOM. An accessible page is eligible, not proven indexed. A CAPTCHA blocking your browser tells you nothing about Googlebot. - Inspect the actual product link on the sample listings. Which element, which
href, whichreltokens, where it finally lands. A rendered anchor can be crawlable even if JavaScript inserted it; a button with anonclickhandler is not a reliable link.nofollow,ugcandsponsoredare hints to Google, not proof of zero value. - Write the evidence down with a date, per product, and never generalise from one listing to the whole directory. Mixed samples stay unknown.
From that evidence the catalog’s scoring script builds a queue. The scoring is deliberately simple: a fit score of 0–3, an audience score of 0–3 that requires an actual observation (not a feeling), and an SEO score that is only non-zero when you have both a confirmed Google result and a crawlable link. Eligibility gates everything: paid-only routes go on hold under a no-spend policy, unknowns go to research, and anything already pending or live goes to follow-up. Value is divided by estimated effort, so a decent directory with a ten-minute form outranks a great one that needs a video, three badges and a company registration.
Two rules from the method are worth repeating because most lists break them. Start with three to five candidates, not fifty. And cap research at five minutes per platform — if you cannot resolve a question in five minutes, it stays “unknown” and the directory stays in the research lane. The zero-spend launch planning guide shows what such a queue looks like in practice.
A short list to start with, by product type
If you want a concrete starting point instead of a method, here is how the categories map to common product types. This is fit, not a promise of results.
- A SaaS with a launch moment: two or three product-launch platforms (Product Hunt plus one or two smaller ones where you can pick your date), one alternative-finder such as SaaSHub or AlternativeTo, and — if you are a company — one B2B review platform. Our page for SaaS founders walks through this.
- An AI tool: the same launch platforms, then three to five AI directories that match your sub-category, then the alternative-finders.
- A developer tool or library: alternative-finders and software catalogs first, launch platforms second, and skip the B2B review sites unless you sell to teams.
- A side project without a company: launch platforms and community launches; mark the company-profile and review platforms as not a fit and move on.
Whatever you pick, the order of operations stays the same: search for an existing listing before creating one, prepare the copy within the field limits, decline every paid upsell, submit only where you have decided to, and record what actually happened.
Where the agent comes in
Nothing above requires an AI agent. It requires discipline and a place to write things down. What an agent changes is the cost of that discipline: reading a playbook, checking three sample listings, filling a form within its limits and recording the outcome is repetitive, rule-bound work that a coding agent with a browser connection does well — and that a founder does badly at eleven at night for the fortieth directory.
That is what LaunchRepo packages: the 346 playbooks, the priority method, a product template your agent reads once, and a tracker that keeps scheduled, submitted, pending review, needs you and live apart. If you want to see how the run looks from your side, the Claude Code walkthrough shows the start prompt and the handoffs. If you are comparing that against doing it by hand, the manual-submission comparison lays out the assumptions openly. And the pricing page has the one number we do publish: what the repository costs, once.
Sources: the LaunchRepo catalog (directories.json, 346 entries and category counts as of 17 September 2026), the catalog’s PRIORITY.md, and the platform list used on this site. No traffic, ranking or authority figures were used because none were measured.