ainative-ax

Score how ready a website is for AI agents, from your terminal. Checks for the signals agents actually look for — llms.txt, agent.json, MCP and OAuth discovery, structured data, API catalogs, and more — and returns a 0–100 score, a letter grade, and a per-check breakdown.

npx ainative-ax audit example.com

Wraps the same scoring engine behind ainative.ax — no install, no API key required for a quick scan.

Usage

npx ainative-ax audit <url> [options]   Run an agent-readiness scan
npx ainative-ax login                   Sign in (stores a session locally)
npx ainative-ax logout                  Clear your stored session
npx ainative-ax history [--json]        List your past scans while signed in
npx ainative-ax --help                  Show this help

Audit options:

--json               Print the raw API response as JSON (no color) —
                      pipeable into jq or a CI step. Exit code unchanged.
--full                Run the full audit (agent simulation + benchmark
                      percentile) instead of the quick scan.
--industry <name>    Industry for the full audit's benchmark comparison
                      (default: saas). Only used with --full.

Exits with status 1 if the scan is incomplete (score below max) or fails, so it can gate CI:

# example GitHub Actions step
- run: npx ainative-ax audit ${{ github.event.repository.html_url }} --json

What it checks

Agent discoverability (llms.txt, agent.json, agent cards), MCP and OAuth discovery, structured data (JSON-LD, Schema.org, OpenGraph), API documentation and catalogs (OpenAPI, RFC 9727), authentication standards, content negotiation, security transparency, and more — the exact check count grows over time. Run a scan or see the full breakdown at ainative.ax.

Full reports, benchmarks, and the leaderboard

This CLI covers the quick scan and full audit. For stored reports, industry benchmarks, remediation guidance, and the public Agentic Web Index leaderboard, visit ainative.ax.

Using it from an AI agent instead?

AX Audit also runs as a Model Context Protocol server at https://ainative.ax/mcp — no separate install. See the main repo README for the MCP tool reference.