# Agent Readiness — Overview

The **[agent-readiness scan](/agent-readiness)** answers one question: can AI coding agents actually use your site? It runs **19 read-only checks** against your public pages — no account created on your systems, nothing changed — and scores four things:

| Category | Weight | The question it answers |
| --- | --- | --- |
| **Access** | 25 | Can an agent fetch anything at all? (robots.txt rules, bot blocking, soft-404s, JS-only pages) |
| **Discovery** | 20 | Can it find your machine entry points? (llms.txt, llms-full.txt, sitemap, docs, agent files) |
| **Comprehension** | 25 | Can it read what it fetches? (markdown availability, quickstart, structure, token budgets, dead links) |
| **Integration** | 30 | Can it act on your product? (OpenAPI, package names, quickstart URLs, auth path, MCP) |

Every check comes back as **pass** (full points), **partial** (half), **fail** (zero), or **unverifiable** (excluded — the score renormalizes so you're not punished for what couldn't be measured). See the **[checks reference](/docs/agent-readiness/checks)** for what each check verifies and how to pass it.

## Grades

| Score | Grade | Meaning |
| --- | --- | --- |
| 90–100 | **A** | Agent-ready |
| 75–89 | **B** | Minor friction |
| 50–74 | **C** | An agent will struggle |
| 25–49 | **D** | An agent will likely fail |
| below 25 | **F** | Effectively invisible to agents |

::callout{type="note"}
**Content sites are scored honestly.** If your site has little for an agent to *integrate* (no API, no packages — say, a blog or a brochure site), the Integration category is skipped entirely and the score renormalizes over the rest. You get an honest "little here for an agent to integrate," not a scary number for missing an API you don't have.
::

## Scan limits

Each scan makes ~40–60 read-only requests against the target, so scans are metered:

- **Without an account:** 1 free scan per day.
- **With a free account:** 5 scans per day — enough to fix and re-check as you go. [Sign up](/sign-up) (no credit card).

## Shareable reports

Every scan gets a permanent URL (`llms-txt.io/agent-readiness/<id>`) you can share — the full scorecard, date-stamped. Scores are a **snapshot of your public surfaces at scan time**, not live: if you've shipped fixes since, re-scan for a fresh report.

## The static scan vs. the Benchmark

The scan is static analysis: what an agent can *see*. It deliberately does not answer whether an agent actually gets from your docs to a working integration — that's execution, and it's what the **[Agent Onboarding Benchmark](/benchmark)** measures: a real coding agent attempts your onboarding (install, auth, first working call) and we report exactly where it fails, how it recovered or didn't, and what to fix first.

## See where sites land

The **[Agent Readiness Index](/agent-readiness-index)** ranks 163 developer-tool and API docs sites on the same 19 checks — useful for calibrating what an A looks like versus a C, and where your site would sit.