How AgentDish works

A cleaner way to browse AI tools and projects.

Submit a public URL. AgentDish checks the page, runs a structured OpenAI review, and publishes only listings that clear the bar.

Review flow

Four checks, no long questionnaire.

1

Send

Submit the product page URL.

2

Scan

Status, metadata, headings, and visible copy.

3

Score

Clear, useful, specific, complete.

4

Publish

Approved listings go live. The rest stays out.

Publishing bar

The score rewards evidence, not launch-page gloss.

Clear30
Useful30
Specific20
Complete20
Clear

Can a human tell what this is? We look for a straight answer to "what does it do?" and "who is it for?" instead of buzzword-dense marketing fluff.

Useful

Proof of life. We reward evidence of a working product: setup steps, usage examples, docs, or live demos. If we can't see how to use it, it doesn't get listed.

Specific

Real tools, not just wrappers. We check for deep integrations, developer APIs, or specialized workflows that solve a non-trivial problem.

Complete

Is there enough here to trust? We look for the basics a visitor needs: pricing, maintenance signals, and enough detail to explain the project without guessing.

Evidence over claims

The page has to show what the product actually does.

No filler

Duplicates, placeholders, broken pages, and spam stay out.

New pages first

Already accepted URLs are skipped so scan budget stays focused on new submissions.