Scroll to the bottom of almost any tag, industry, region, area or VC page on APIs.io and you will find a section titled Cohort brief, marked Auto-generated. There are 1,419 of them: 985 on tag pages, 216 on regions, 96 on venture funds, 79 on industries and 43 on areas. They are built from the catalog build of 26 September, which covered 27,598 providers.
What a brief is
A cohort is any roster the catalog can name: a tag, an industry, a region, an area, or a fund’s portfolio. Each brief reports the same things. It gives the mean Kin Score and mean Agent Readiness against the catalog baseline (23 and 12.2), the spread, a single bar showing the band split, the facet averages against the catalog, and a table of adoption rates for MCP servers, OAuth scopes, Arazzo workflows and governance rules. It ends with two leaderboards, one ranked by composite score and one ranked by agent readiness.
Two design decisions make the briefs worth trusting. First, the page cannot say anything the API would not. The generator imports the search Lambda’s own cohort module and calls the same functions that serve /v1/cohorts/{kind}/{slug} and the MCP tools find_cohorts, get_cohort and cohort_rankings. Second, a brief never calls its roster a “market”. Most briefs are tags, and the tag vocabulary mixes real markets like payments with resource nouns like addresses. So the lede says “the 623 providers in the APIs.io catalog classified under Logistics” and stops there.
Why the second leaderboard matters
On the Logistics page, AfterShip tops both lists. But ranks two through four for agent readiness are Optilogic (54.9), Affinda (54.2) and Loadsmart (51.0), and none of them is in the composite top ten. If you are choosing an API for an agent to drive rather than for a developer to read, that second list is the one to use.
The briefs are also honest about the enrichment we have and have not done. Members flagged as never enriched are left out of the averages. The cohort data behind each brief also counts how many scored members had under half of their catalog-attributable points satisfied: 548 of 623 in Logistics. A low mean can be partly our backlog. The API says so. The rendered brief does not show it yet, and it should.
The sentence that is wrong
The coverage disclosure at the bottom has two branches. When a roster declares more members than resolve, it says so. When every member resolves, it prints “All N members of this roster resolve to a scored provider in the catalog”. Resolving and being scored are not the same thing. Logistics resolves all 670 of its members and scores 623, so its brief claims all 670 are scored. The same thing happens on 970 of the 1,419 briefs. The numbers above that line are correct. Only the sentence overclaims. The fix, which is ours to make, is to compare against the scored count.
The Kleiner Perkins page shows why the disclosure matters. Its Portfolio Strength card reports a composite average of 39.3 across 87 publishing companies. The cohort brief a little further down reports a mean of 26 across 165 scored companies. Both are correct for different rosters, and only the disclosure tells a reader which is which.
Takeaway
The cohort brief turns 1,419 API responses into something a person can read, without letting the page drift from the API. Fix the one sentence and it is the best-disclosed number on APIs.io. Start with Logistics and scroll to the bottom.