Agents are reading your documentation. Whether it's Cursor pulling context for a developer, Claude answering a support question, or an autonomous agent evaluating your API, your docs have a new audience that never tells you if it got what it needed from your site.
To be agent-ready, your documentation must be structured so agents can find it, understand it, and use it. That means clear entry points, well-formed content, consistent terminology, and the right signals in the right places.
Agent score tests your documentation against 29 checks and tells you exactly how ready it is for agents.
Coding agents are rapidly becoming the primary surface for developer experience. To tackle anything meaningfully complex, agents require detailed, trustworthy context. For APIs and SDKs, that means documentation. When docs are accurate, well-structured, and agent-accessible, devs can pull off nearly hands-free integrations, saving huge amounts of time and effort. It's early days, but in customer conversations we're already seeing this emerge as a competitive differentiator.
— Mo King, Lead DX Engineer at Together AI
Enter your documentation site URL at mintlify.com/score. You'll get a score and an explanation of each category in seconds.
You'll see what your site does well and where you can improve.
Keep in mind, it's a snapshot. Your score reflects the current state of your documentation. It changes as your docs change.
A low score isn't a failure. Most documentation wasn't written with agents in mind, because agents weren't a primary audience until recently. Use your score to guide where you take your documentation rather than judge it's current state. Documentation maintenance will always be an ongoing process.
Agent score is built on the Agent-Friendly Documentation Spec, an open standard that defines what good documentation looks like for agents. It runs checks across seven categories:
The spec covers the foundations very well. We added checks on top based on how we observe agents using documentation:
These additions reflect what we're seeing matter most right now, and how we expect agent-mediated documentation use to evolve.
We built an internal tool with proprietary scoring. It worked well, but building on top of a closed standard didn't make sense.
Agent-readiness isn't a problem anyone will solve alone. Standards like llms.txt only work when they're widely adopted. The signals agents rely on only matter when they're consistent. The ecosystem only improves when the whole documentation environment iterates together.
The Agent-Friendly Documentation Spec is an open standard, with a companion open source library that implements all the checks.
We're building with the community and want to encourage others to contribute to the standard and the tooling.
Agent score is built on the spec. The checks are transparent. We're highlighting additional patterns we believe will improve documentation for all of our users.
Writing documentation for agents is a new discipline. We're all learning, and everyone improves with collaboration. More voices will help the standards and best practices develop faster.
If you think we're missing something, we want to hear from you. If you disagree with how we've implemented a check, let us know. If you've found something in your own docs that correlates with good or bad agent behavior, that's exactly the kind of signal that can help everyone who maintains documentation.
Score your docs at mintlify.com/score.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。