How it works

This directory is built by an automated discovery engine, not by hand.

What this is

AI Tools Discovery is an independent discovery engine. Once a day it queries a set of public sources, works out which of the results are genuinely usable AI tools, resolves each one to its own website, and publishes what it can verify. Nothing here is a paid placement, and no listing is written by hand unless an editor has explicitly reviewed it.

Where the data comes from

Tools are discovered from the public APIs and feeds below. Each tool page lists exactly which of these it was found in, with a link back to the original record.

  • GitHub
  • Hugging Face
  • Hacker News
  • RSS
  • Google News
  • PyPI
  • npm
  • Docker Hub
  • OpenAlex
  • GDELT
  • YouTube
  • Product Hunt

The pipeline

  1. Discovery

    Each source adapter fetches only what has changed since the last successful run.

  2. Normalisation

    Every result is mapped onto one common shape with a stable content hash.

  3. Validation

    Heuristics reject tutorials, awesome-lists, courses, datasets and dead projects.

  4. Deduplication

    Canonical URL, domain, external id, repository and fuzzy name matching.

  5. URL resolution

    The tool's own website is resolved; a news article is never used as one.

  6. Content extraction

    Selected official websites are read once each, within a strict budget.

  7. Classification

    A language model decides whether the candidate really is an AI tool.

  8. Enrichment

    Category, pricing, features and use cases are extracted and schema-validated.

  9. Scoring

    Quality, popularity, freshness, discovery and trending scores are computed.

  10. Publishing

    Only records that clear the confidence bar go live; the rest await review.

What we will not do

We never invent facts. Pricing, features, integrations and ratings are only shown when a source states them. Where a pricing model cannot be confirmed, the tool is labelled Unknown rather than guessed at.

We never republish articles. Only concise metadata and generated summaries are stored, and the original source URL is always kept.

We never fabricate statistics. Every count on this site is a live query against the database.

Current state

Tools
759
Categories
28
Open source
614
With an API
106

Last successful discovery run 37m ago.

Corrections

If a listing is wrong, out of date, or should not be here, tell us via the submission form and an editor will review it.