With 46 tools and 470 MCP servers, there are millions of possible pairs. We publish 21 comparisons. The constraint is not effort — it is that most pairs do not represent a decision anyone makes.
The test a pair has to pass
**Would someone genuinely choose between them, once?** Cursor and Windsurf are a decision. Postgres and Discord are not, even though both are MCP servers.
**Is there a real trade-off?** A comparison where one option is better on every axis is a link, not a page. The pairs we keep have a genuine cost on both sides: Jira's process against Linear's speed, Sentry's depth on one error against Datadog's view of the system.
**Can we say something that is not in both marketing pages?** If the answer is 'both are great, it depends', the page should not exist.
Why the bottom line comes first
Every comparison opens with a one-sentence decision. If the honest sentence is 'whichever you already pay for' — as with the two email connectors — we write that, and the rest of the page explains why the question is not worth more of your time.
A comparison that refuses to conclude is a comparison that wasted a click. The categories where a tie is real are exactly the ones where saying so is most useful.
What gets compared next
The order is demand-led: comparisons people search for, then pairs the alternatives pages keep implying. If a page produces a question we cannot answer, the answer usually becomes the next comparison.
What we will not do
We will not compare two entries where one is listed and the other reviewed. A generated description against a checked one is not a comparison, it is an advertisement for the reviewed tier — and it would make the listed tier look bad for the wrong reason.