How we test
Stop paying for software, but not blindly. Every tool page says how far we have proven the tool, what the catch is, and what you would save. Here is exactly how each of those is decided.
Proof levels
Every tool carries one of four badges. Each is earned by a fixed rule, never by opinion and never by payment. The highest level a tool has earned is the one shown.
| Badge | How it is earned | Tools |
|---|---|---|
| Used | We did real jobs with it, start to finish, in a published deep-dive video. | 1 |
| Run | We ran it on our own server and recorded the demo on this page. | 10 |
| Seen | We installed it on our server and tried a real job on camera in one of our videos. | 9 |
| Read | We read its README and licence and ran our automatic checks. We have not run it yet. | 1,276 |
Run means a demo recorded on our server. We start the tool in Docker on our own server, follow a short written script with the tool's real buttons and fields, and record the screen. Nothing is mocked up: what you see is the real tool doing the real job. Waiting is sped up, and every demo says the date it was recorded.
Demos so far: omni-tools, memos, linkding, Ghost, docuseal, changedetection.io, uptime-kuma, Stirling-PDF, miniPaint, excalidraw, AudioMass.
The saving
- We only show a saving where we have the paid app's price from its own pricing page (or a named source), with the date we checked it.
- The yearly figure is that monthly price times twelve, in the currency the app charges in. We never convert currencies or guess.
- It is the entry price for the plan that does the job. If the paid app has a free plan that would do for most people, we say so next to the saving, and if most people would not pay at all, we count nothing.
- Your plan's total adds up only the apps with a checked price.
The catch
Every tool gets one line saying what to check first, chosen by rule in this order: an unusual star pattern, a licence that is not open source or not stated, the first problem we found in its documentation, the AGPL's sharing rule, needing a graphics card, needing your own server, being under two months old, then a technical setup. If none apply, we say so.
Verdicts
- Check first: an unusual star pattern (see star watch).
- Check the licence: source-available or no clear licence.
- New and promising: under two months old.
- Worth a look: established, with a few things to check.
- Ready to use: established, openly licensed and no red flags in its documentation.
Where we look
GitHub search, Hacker News, YouTube creators, curated lists and the official MCP Registry, every day.
What we check
- Licence: read from the project's licence file and sorted into permissive, weak copyleft, copyleft, source-available or unclear. This decides whether you can build a paid product on it.
- What it needs to run: your computer, a small server, a GPU, a phone or a browser.
- Growth: stars gained per day. Patterns that often go with bought stars are flagged on star watch and kept out of every ranking.
- Safety: we do not list malware, exploits, cracking tools, or anything built to break other services' rules.
How summaries are written
Summaries are drafted by an AI model from each project's README. Licence, safety and star-pattern checks are automatic rules. Some summaries may contain mistakes, so please tell us if you find one. "Could replace" is our opinion of the overlap with a paid product, not a claim that every feature matches.
Who we are
An independent project. We are not affiliated with any of the projects listed and we do not accept payment for listings. We also make a few paid kits of our own; they are labelled as ours, sold in the Shop, and never ranked inside the lists. Sponsors are labelled too, and money never buys a verdict.