You are weighing platforms for a new serverless project, and you open the Serverless page on APIs.io. It says 141 related providers. The overview says "30 providers on the network work in this area." You scroll, count the cards, and find 30. You start building a shortlist.
Here is what the page does not tell you. The 30 are a display cap, taken from the top. The page's own overview sentence is a build bug. The 30 have a median Kin Score of 67.5, and none sits below the strong band. The 134 scored records that can be resolved from the area's membership list have a median of 44.4, and 57 of them are thin, emerging or minimal. The source names 135 slugs, and one is not a provider record, so these counts cover 134 of the 141 the page claims.
Split by band, the fuller picture looks like this: 17 exemplar, 30 strong, 30 developing, 30 thin, 14 emerging and 13 minimal. At the top you would find Amazon DynamoDB at 80.2, Appwrite at 78.3, Cloudflare at 75.6 and Netlify at 75.0. Further down sit Upstash at 37.2, Fly.io at 35.0 and OpenFaaS at 34.4. Near the bottom are MotherDuck at 20.2, Kinvey at 7.0 and Fauna at 0.0.
Some of what you would miss matters for your choice. AWS carries a lot of the top: 17 of the 134 are Amazon or AWS records, with a median of 63.1, while everyone else has a median of 42.0. AWS gets a record per service, so it holds a large share of the upper bands. Google's records sit lower. Cloud Functions and Cloud Run are at 54.3 and Workflows at 51.5, and none of the three is in the rendered 30. The post is careful about this: that gap is a difference in what the record shows per service, not a verdict on the runtimes.
If you care about agents calling your platform, the order changes again. Cloudflare leads agent readiness at 60.6, then InsForge at 55.0, even though InsForge's Kin Score is only 58.1. Blaxel is the sharper case, at 33.7 on the Kin Score and 47.1 for agents, which puts a thin record seventh in the area. And 40 of the 134 list an MCP server.
Then there is the graveyard. Nine providers score 0.0: Binaris, Bit.io, Epsagon, PiCloud, StackPath, Suborbital, Thundra, Fusebit and Fauna. Their records describe them in the past tense, and Fauna's says the hosted service shut down in 2025. A market map that drops them loses the history, but you should not read them as current choices.
Now the turn. This is not a story about a market that hides its weak players. It is a catalog reporting on itself. The post says plainly that the 30 is the catalog's own display cap and that the overview sentence is a build bug. It also lists other things it has to clean up. The page's cohort brief splits its 30 as 10 exemplar and 20 strong, while the band pills on the same page show 18 and 12. Azure Functions has two records, Microsoft Azure Functions at 62.1 and Azure Function Apps at 43.8, and both descriptions cover the same service. And AWS Glue alone holds 203 of the 513 APIs those 134 records count, which skews any count by API.
So what do you take away? Read the page as a shortlist and not as the market. The sample shows exemplar and strong platforms only, while the wider record is a broad middle with a long tail. If a provider you care about is missing from the 30, that is not a verdict on it. Look it up directly, check whether it has duplicate records, and read the band and the agent score separately. The catalog says it should render the full membership or say plainly that the list is a sample, and until it does, you now know to treat it as one.
Read the original: https://apis.io/2026/11/02/serverless-page-shows-30-of-141-providers-and-the-sample-flatters-the-market-by-23-points/