Launching today
Jentic API Scorecard is a free, open-source CLI and web UI that grades any OpenAPI spec for AI-readiness across six dimensions: compliance, developer experience, agent readiness, usability, security, and discoverability. AI agents are already doing remarkable things with APIs. The question is whether they can do them with yours. Ensure your APIs are ready for agents. One command returns a single score and exactly what to fix. Free and open under Apache 2.0.
Hi Product Hunt 👋
I'm Frank, Head of Enterprise Architecture at Jentic. AI agents are getting genuinely good at calling APIs to get real work done, and the best teams are already building remarkable things on top of them. Every company now has an AI strategy, and a lot of those strategies quietly depend on agents being able to use the company's existing APIs. That is where the gap shows up.
The reason is straightforward. APIs were designed with human developers in mind. When a developer hits a vague description, a missing example, or an inconsistent response, they fill in the gaps. They read the source, ask a colleague, or make an informed guess and move on. Agents are perfectly capable of operating an API, but only when the quality is built in. When a description is ambiguous or behaviour is unpredictable, an agent has nothing to fall back on, so it routes around your API and uses one that is easier to work with. You never see the traffic, the integrations, or the use cases you missed. The distance between an API that passes a standard linter and one an agent will actually choose is much larger than most teams realise, and there has been no consistent way to measure it.
We built the Scorecard to put a number on that distance. It started as an internal tool for assessing the thousands of APIs in our own catalogue, and it became clear that every team building or exposing an API has the same blind spot.
What the Scorecard does
It scores any OpenAPI-described API across six dimensions:
- Foundational Compliance: is the description technically correct?
- Developer Experience & Jentic Compatibility: is it clear and usable?
- AI-Readiness & Agent Experience: can an agent interpret it?
- Agent Usability: can an agent operate it reliably?
- Security & Governance: are the right guardrails in place?
- AI Discoverability: can an agent find and execute against it unaided?
You get one headline grade and a per-dimension breakdown of what to improve.
As a CLI it runs from your terminal with one command, no browser required:
or it can be used as via the web UI here:
https://jentic.com/scorecard
Why it matters We think AI-readiness is heading the same way as security testing. A decade ago you ran a security audit before launch and hoped for the best. Today those checks run on every commit, and nobody questions it. The Scorecard is built to slot into the same place in your workflow. Run it once to see where you stand, then wire it into CI so every change is scored automatically. Over time you get a trend line showing your APIs becoming more agent-ready with each release, and you stop shipping changes that quietly regress it. The scoring framework is published under Apache 2.0 and was developed with input from the OpenAPI Initiative community, including a member of its Business Governance Board and an OpenAPI Ambassador, both now at Jentic. We want this to be a shared, open standard rather than a closed metric, so it is fully open to inspection, contribution, and extension. The CLI and web UI are completely free. We would genuinely like you to run it against your own API and tell us the score you got, and whether it matched what you expected. Tell us which APIs we should test next, and where the framework still has gaps.
@jentic @frankkilcommins Congrats on the launch Frank. What are you finding irl?
@zolani_matebese Thanks!
Real world results have been eye-opening. After scoring 6,000+ APIs, the pattern is clear: there's a huge amount of low-hanging fruit sitting on the table. The issues we see most often aren't exotic edge cases, they're fundamentals:
Invalid or unparsable OpenAPI docs
Missing path parameters and broken $ref chains
Invalid or missing examples
Inconsistent naming and unclear verb usage
Vague or generic descriptions that mean nothing to an agent
Missing server definitions or wrong environments
Auth info living outside the OpenAPI document entirely
Stale, outdated specs that no longer match the actual API
The mix varies by team, but almost every API has at least two or three of these. And the tricky part is that many of these APIs pass validity tests — they're technically "valid" but practically unusable by an agent.
That's the core problem we set out to solve: you can't improve what you can't measure. The Scorecard tells you whether an agent will actually succeed with it. Those are very different questions.
Once teams have a score and a per-dimension breakdown, the path to improvement becomes obvious. That's what we been hearing initially too — not "this is hard to fix" but "we didn't know this was broken."
Would love to hear what scores people get when they run it against their own APIs!