Agentfix Oss Scanner

agentfix-mini-scanner

[](https://www.npmjs.com/package/agentfix-mini-scanner) [](https://www.npmjs.com/package/agentfix-mini-scanner) [](https://nodejs.org) [](LICENSE)

A tiny, zero-dependency, MIT-licensed scanner that checks whether your website is discoverable and usable by modern AI agents (ChatGPT, Claude, Perplexity, Gemini, MCP clients).

> Bad score? Skip the spec rabbit-hole. AgentFix Pack > scans your site across the full AI-agent readiness checklist and emails you a ready-to-install ZIP > with your real llms.txt, schema, A2A agent-card, MCP server-card, and > install guides for WordPress / Webflow / Tilda / Shopify / cPanel. One-time > $1 Mini · $29 Pack · $99 Pro (drift monitoring). No subscription. > > → Scan your site at agentfix.pro

It runs the discovery-critical subset of signals the commercial AgentFix scanner uses (see the full AI-agent readiness checklist). If your site passes all twelve here you're ahead of ~95% of the web. If it fails most, the full audit + ready-to-install fix pack at agentfix.pro closes the gap in minutes.

See it visually - the AgentFix Workspace iceberg

For a free interactive visualisation of what AI agents look for, paste any URL into workspace.agentfix.pro. The Agent Lens window renders your site as an iceberg: the tip is what classic Google indexers see, just under the waterline is what AI agents reach for (llms.txt, schema, robots), deeper is the agent-native protocol layer (A2A, MCP, OpenAPI), and the depths are signals nobody ships yet (ACP/UCP commerce, streaming corpus). Same 12 checks as this CLI, rendered side-by-side with Google vs AI vs Optimal.

$ npx agentfix-mini-scanner example.com

AgentFix Mini Scanner - discovery-critical subset
Target: https://example.com

  PASS  /llms.txt accessible
        842 bytes
  FAIL  /llms-full.txt accessible
        /llms-full.txt missing
        → Add /llms-full.txt with the long-form text content of your site.
  PASS  robots.txt allows AI crawlers
        Mentions: gptbot, claudebot
  FAIL  schema.org Organization JSON-LD
        No JSON-LD blocks found in homepage HTML
        → Embed a  Organization block.
  ...

Score: 50% (Grade D) - 6/12 passed

What's new in 1.1

Four additional checks added:

• schema.org WebSite + SearchAction - sitelinks search box
• /openapi.json - OpenAPI 3.x service-desc for agents
• /.well-known/api-catalog - RFC 9727 linkset+json
• /.well-known/oauth-protected-resource - RFC 9728 auth metadata

8 → 12 signals. Same zero dependencies.

Install

npm install -g agentfix-mini-scanner

or run once without installing:

npx agentfix-mini-scanner https://your-site.com

Node 18+ required (built on the global fetch).

What it checks

| # | Check | Category | |---|---|---| | 1 | /llms.txt accessible | discovery | | 2 | /llms-full.txt accessible | discovery | | 3 | robots.txt explicitly allows AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) | aipolicy | | 4 | schema.org Organization JSON-LD on homepage | schema | | 5 | schema.org WebSite + SearchAction JSON-LD on homepage | schema | | 6 | /.well-known/mcp/server-card.json (Model Context Protocol) | agentprotocols | | 7 | /.well-known/agent-card.json (Agent-to-Agent / A2A) | agentprotocols | | 8 | sitemap.xml accessible | discovery | | 9 | Link: header advertising service-desc / api-catalog / agent-card | agentprotocols | | 10 | /openapi.json (OpenAPI 3.x) | agentprotocols | | 11 | /.well-known/api-catalog (RFC 9727 linkset+json) | agentprotocols | | 12 | /.well-known/oauth-protected-resource (RFC 9728) | agent_protocols |

Programmatic use

import { scan } from "agentfix-mini-scanner";

const report = await scan("https://acme.com");
console.log(report.summary);
//  { pass: 6, fail: 6, skip: 0, total: 12, score: 50, grade: "D" }

for (const c of report.checks) {
  console.log(c.key, c.status, c.detail);
}

scan(url) always resolves; individual checks catch their own errors and report status: "fail" with the cause in detail. Network timeout per check is 8 seconds.

What this scanner does NOT check (the rest of the checklist)

The commercial AgentFix scanner additionally evaluates: declarative shadow DOM accessibility, MCP tool input-schema validity, A2A skill descriptors, agent-installable scripts (.well-known/install), per-page meta tags, Article/FAQ/Product schema, RFC 8288 markdown content-negotiation, RSS/Atom feed exposure, anti-bot WAF heuristics, heading hierarchy, image alt coverage, anchor-tag density, Cloudflare Bot Verifications, SSR pre-render readiness, ACP/UCP commerce stubs, runtime ARIA fixes, and several more (see agentfix.pro/docs).

Why open-source the basics?

Two reasons:

1. We think every site owner should be able to check at least these twelve

signals for free, forever. They're the discovery foundation; without them AI agents can't even find your content.

1. If you run this and your score is bad, the $29 fix pack at agentfix.pro

ships you a personalised ZIP that closes the gaps in a few minutes. The OSS scanner is the on-ramp to the paid product, and we're upfront about that.

When to use what

| Goal | Tool | |---|---| | Quick CI check / curiosity / dev workflow | This CLI (free, MIT, npx) | | You see your score is low and want it fixed today | AgentFix Pack $29 - ZIP with your llms.txt, schema, agent-card, server-card, install guides | | You only need llms.txt to start | AgentFix Mini $1 | | You want monthly drift monitoring + diff emails | AgentFix Pro $99 | | You want to see what AI agents look for, visually | workspace.agentfix.pro (free, interactive iceberg) |

Changelog

• 1.1.4 (2026-07-10) - dropped hardcoded "12 of 33 signals" phrasing;

README, help banner and package description now point to agentfix.pro/methodology as the single source of truth for the checklist (currently 34 signals across 7 categories, and the number will keep evolving). CLI checks unchanged.

• 1.1.3 (2026-07-05) - README polish + npm/Node badges. Package keywords

expanded (MCP, A2A, GEO, OpenAPI, RFC 9727/9728, GPTBot, ClaudeBot) so npm and GitHub search surface the scanner for people looking to fix llms.txt, agent-card, server-card and the rest of the agent-readiness stack.

• 1.1.2 (2026-06-27) - README pivoted to lead with the AgentFix Pack pitch
• when-to-use-what matrix. Workspace iceberg companion link prominent.
• 1.1.1 (2026-06-27) - added Workspace iceberg link to README. Live

visualisation companion at workspace.agentfix.pro.

• 1.1.0 (2026-06-26) - added WebSite schema, openapi.json, api-catalog,

oauth-protected-resource checks. 8 -> 12 signals. Fixed GitHub repository URL.

• 1.0.0 (2026-06-19) - initial release with 8 checks.

Support development

If this scanner helps you, buy us a coffee on Ko-fi ☕. Every coffee funds the next check; we ship additional signals from the commercial scanner one OSS release at a time.

[](https://ko-fi.com/agentfix)

License

MIT - see [LICENSE](LICENSE).

Contributing

PRs welcome for the twelve checks above. New checks belong in the commercial scanner so we can validate them against real buyer sites first; open an issue if you have a candidate signal worth promoting to the OSS version.

Bug? Open a GitHub issue or email support@agentfix.pro.

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Virt92
• Source: Virt92/agentfix-oss-scanner
• License: MIT
• Homepage: https://agentfix.pro

Install and usage instructions live in the source repository linked above.