Imagine you run product at a company that sells an AI agent to customer support teams. A prospect asks a plain question: can our agents work with yours? You open a catalog that tracks how readable companies are to software, and you look up your own row. It is not what you hoped for.
That is roughly where the APIs.io customer support vertical leaves you. The catalog holds 320 providers and 3,237 APIs in Customer Support & Service. Of the 318 providers that are scored, the median Kin Score is 31.4 and the median Agent Readiness is 19.6. Those are the numbers to hold in your head as the baseline.
The sharpest result is the one about you. Companies whose product is an AI agent can sit near the floor on being usable by one. Decagon scores 0.6 on Agent Readiness, Cresta 2.5 and Sierra 4.1. All three are on what the catalog calls a split basis, which means the record is built from the catalog cutting up what they publish. It is not built from a contract they released themselves.
That split matters, and the post says so directly. The 123 providers on a published basis have a median Kin Score of 51.0 and an Agent Readiness of 30.8. The 195 scored on a split basis sit at 19.3 and 2.5. Half of the split group is close to invisible to an agent in the record. The post is careful about what that means: it says what the catalog could read, not what the products can do.
Now look at who does well. You might expect the household names. The highest agent score belongs to Comms, a customer messaging API, at 66.1. Simple AI follows at 62.1, then Rulebase at 61.2, ClosedLoop at 59.4 and Brainfish at 59.2. Twilio leads the Kin Score at 86.0 but ranks 20th on agents. HubSpot is second at 84.2 and ranks 28th at 43.7. Zendesk is third among providers tagged Help Desk, yet its 33.3 ranks 60th, behind companies with a fraction of its 150 APIs. Nineteen members score higher on agents than on the Kin Score.
The post then turns on its own record, and this is the part worth your attention. The vertical is built from tags, so it keeps companies whose own records say they are gone. Twenty-five members carry a Defunct or Acquired tag, with a median Kin Score of 2.2. Parature, Get Satisfaction and Witness Systems are among them. Nine members score exactly zero, and Cogito and Remedy Corporation have no score at all. They hold down the median, which is why 31.4 understates the live market. The catalog says removing them from scored cohorts is its decision to make, and it has not yet made it. So you should read the headline median with that admission beside it.
Back to your prospect. Does selling agents mean being ready for them? In this data, no. The 61 members tagged AI Agents have a median of 32.4 and 19.1, below the vertical's agent median of 19.6. Regal, at 76.0 and 50.0, shows the two can go together, so this is not a law of nature.
If you are buying rather than selling, the post gives you somewhere to begin. Start with Twilio or HubSpot for breadth with a contract behind it. Ada is the strongest record among providers tagged Help Desk. For an agent-first build, read Comms, ClosedLoop and Brainfish before the larger suites, and check what each actually exposes.
If you are selling, the takeaway is blunter. Publish a contract your buyers' agents can read. The post says Decagon and Sierra could move further with one OpenAPI document than almost anyone in the table. You walk away knowing the gap is partly about what you publish, not only about what you built, and that the catalog is still cleaning up its own cohort.
Read the original: https://apis.io/2026/11/06/customer-support-vendors-selling-ai-agents-score-near-zero-on-agent-readiness/