# How we score listings

We only count what we can check. If a signal is missing, we say "not enough data." We don't invent popularity, and we don't fill gaps with estimates.

HTML: https://agentready-tools.com/methodology
Live ratings JSON: https://agentready-tools.com/api/ratings

## Agent-readiness

This score is a checklist, not a vibe. Each item is yes, no, or unknown:

- Machine-readable interface (MCP, OpenAPI, or a structured interface we have confirmed) — 25 points
- Auth type is documented — 15
- Pay-per-use pricing is stated — 20
- Agent controls such as limits or scopes — 15
- Tool annotations, for example a read-only hint — 10
- The official URL responded in a recent liveness check — 15

Unknown items add nothing and are not treated as failures. A numeric score appears only when at least three items have been verified. Until then the listing shows the checklist and the label "not enough data."

Most directory listings will sit in that state for a while. We have verified more of the checklist on the tools we built ourselves, because we can stand behind those facts. We do not guess auth, pricing, or controls for a third-party tool just to produce a number.

## Popularity

Popularity is a separate measure. It is not mixed into agent-readiness. We show each source on its own:

- Outbound /go clicks on this site, last 7 days, with agent clicks labelled separately
- MCP `get_tool` lookups, last 7 days
- Views of a tool page on this site, last 7 days
- Public GitHub stars, when the official URL is a GitHub repository we could read
- Public npm weekly downloads, when a package name is on file

Test traffic we can identify is excluded. If a source is missing, it stays labelled as not enough data. A zero is only shown when we actually measured zero in a working log or public API. The [trending](https://agentready-tools.com/trending) page uses the same first-party logs. If there isn't enough real traffic, it falls back to an editorial shortlist and says so.

## Scan report badges

A listing may show a factual badge such as "Scan report submitted · Snyk · 2026-10-01 · 1.2.3," with a link to the report. That badge is informational only. It is not part of the agent-readiness score, and it has no weight in popularity.

We do not word these badges as secure, verified safe, certified, or approved. If the scanned version no longer matches a later version we have on file, or if the report is older than 180 days, the badge is marked stale or hidden.

Scan badges, and any reports, files, or URLs associated with them, reflect material supplied by third parties. AgentReady Tools does not conduct, review, verify, audit, or endorse such scans or reports, and a scan badge is not a representation, warranty, certification, approval, or assurance of security, safety, quality, or fitness for any purpose. Reliance on a scan badge or on any third-party scan report is at the sole risk of the reader or operator. The [Terms](https://agentready-tools.com/terms) govern the Service, including scan badges.

## Where the numbers appear

Both scores, and a scan badge when one is on file, are included in [tools.json](https://agentready-tools.com/tools.json), in per-tool JSON records, in the MCP `get_tool` output, and on tool pages. Live snapshots are also at [/api/ratings](https://agentready-tools.com/api/ratings).
