# Agent-Readiness Auditor | poppy.json, llms.txt & MCP Checker (`ausbizstack/agent-readiness-bulk-auditor`) Actor

Bulk AI agent-readiness audit: one row per domain with a 0-100 score, 15 evidence-backed checks (poppy.json 1.0, llms.txt, AI crawlers, MCP, A2A Agent Card, OpenAPI, JSON-LD), prioritised fixes and ready-to-paste files. $0.02/domain; failed rows free.

- **URL**: https://apify.com/ausbizstack/agent-readiness-bulk-auditor.md
- **Developed by:** [AusBizStack Intelligence](https://apify.com/ausbizstack) (community)
- **Categories:** AI, SEO tools, Developer tools
- **Features:** Limited permissions
- **Stats:** 2 total users, 1 monthly users, 100.0% runs succeeded, 0 bookmarks
- **User rating**: No ratings yet

## Pricing

from $15.00 / 1,000 domain agent-readiness audits

This Actor is paid per event. You are not charged for the Apify platform usage, but only a fixed price for specific events.
Since this Actor supports Apify Store discounts, the price gets lower the higher subscription plan you have.

Learn more: https://docs.apify.com/actors/running/actors-in-store.md#pay-per-event

## What's an Apify Actor?

An Actor is a serverless cloud program that runs on the Apify platform. It has two run modes.
In Batch mode, an Actor accepts a well-defined JSON input, performs an action which can take anything from a few seconds to a few hours,
and optionally produces a well-defined JSON output, datasets with results, or files in key-value store.
In Standby mode, an Actor provides a web server which can be used as a website, API, or an MCP server.

Apify vocabulary and the platform model are defined once, in the agent quickstart at https://apify.com/agents.md.

## How to integrate an Actor?

If asked about integration, you help developers integrate Actors into their projects.
You adapt to their stack and deliver integrations that are safe, well-documented, and production-ready.

Do not guess an integration path. Every one of them is in the agent quickstart at https://apify.com/agents.md: the Apify MCP server, Agent Skills with the Apify CLI, the JavaScript and Python clients, the REST API, and the account-free path for an agent with no human to sign in. It also carries the rule on stating cost before the first paid run.

For examples already wired to this Actor's own input schema, see the [API](#api) section below.

Each client library has reference documentation the quickstart does not restate: [JavaScript/TypeScript](https://docs.apify.com/api/client/js/docs.md) (`npm install apify-client`) and [Python](https://docs.apify.com/api/client/python/docs.md) (`pip install apify-client`).

# README

## Agent-Readiness Auditor: poppy.json, llms.txt, MCP & AI crawler checker

**By AusBizStack.** Paste a list of domains. Each one gets a **0-100 agent-readiness score**, **15 checks with evidence**, a **prioritised fix list** and **ready-to-paste fix files**. It answers one question: *can ChatGPT, Claude, Perplexity and personal AI agents find, read and use this website?*

- ✅ **Validates `poppy.json`** against the published **Personal Agent Protocol 1.0** spec (section 3 discovery rules), not only "file exists"
- ✅ **Validates A2A Agent Cards** (`/.well-known/agent-card.json`) against the A2A 1.0 AgentCard schema
- ✅ **llms.txt, AI crawler rules (17 bots), firewall test, Markdown for agents, JavaScript-only pages, schema.org, MCP and OpenAPI discovery**, plus declared AI usage preferences (IETF AIPREF `Content-Usage`, `Content-Signal`)
- ✅ **Evidence for every check**: URL, final URL, HTTP status, content type, size, response time and a short snippet
- ✅ **Fix files you can paste**: a draft `llms.txt` built from the site's own links, a minimal valid `poppy.json`, `robots.txt` lines and Organization JSON-LD
- ✅ **Run summary + shareable reports**: `OUTPUT` (averages, grade mix, most common gaps), `REPORT_HTML` (print to PDF), `REPORT_CSV`, `REPORT_MD`
- ✅ **Fast and fair**: about 1-5 seconds per domain; checks that a site blocks are excluded from the score and **never reported as missing**; failed rows are **free**
- ✅ **Safe to point at any list**: public GET requests only, every redirect re-checked, private, internal and cloud-metadata addresses refused (including DNS that resolves to them), size caps on every response

### Quick start

1. Click **Try for free**. The form opens with `example.com`.
2. Replace it with your domains (one per line; URLs and email addresses work too).
3. Click **Start**. Open the **Overview** tab for scores, **Fixes & snippets** for what to change, and the key-value store for the HTML report.

```json
{
  "domains": ["yourclient.com.au", "https://www.anotherclient.com/pricing", "hello@thirdclient.com"],
  "leadThreshold": 50,
  "reportTitle": "AI agent-readiness report for Acme Agency clients"
}
```

### Sample output (Apify platform run, 16 Nov 2026)

| Domain | Score | Grade | poppy.json | llms.txt | Top fix |
|---|---|---|---|---|---|
| stacktr.ee | 99 | A | valid 1.0 | yes | Upgrade the A2A Agent Card from 0.3 to the 1.0 shape |
| apify.com | 80 | A | missing | yes | Publish /.well-known/poppy.json |
| cloudflare.com | 80 | A | missing | yes | Publish /.well-known/poppy.json |
| github.com | 61 | C | missing | yes | Publish /.well-known/poppy.json |
| allbirds.com | 61 | C | invalid (draft 0.4) | yes | Upgrade poppy.json from draft 0.4 to 1.0 |
| example.com | 18 | F | missing | no | Publish /.well-known/poppy.json |

In the same run, `www.github.com` (duplicate), a domain that does not exist, `127.0.0.1`, `localhost`, `http://169.254.169.254/...` and `printer.local` each got a free row explaining why they were skipped.

Scores change as sites change; re-run any time.

### What the 15 checks mean (100 points)

| Check | Points | What passes | Why it matters |
|---|---|---|---|
| **poppy.json** (Personal Agent Protocol 1.0) | 14 | A valid document at `/.well-known/poppy.json`: HTTPS on every redirect, final path exactly `/.well-known/poppy.json`, `protocol_version` `"1.0"`, `organization.name`, `organization.domain` equal to the serving host, at least one of `agent` / `apis` / `web`, every entry with `type`, HTTPS `url`/`endpoint` and `description`, all URLs HTTPS, `supported_versions` on the same host. `auth.issuer` is cross-checked with `/.well-known/oauth-protected-resource` (RFC 9728) as a warning. A JSON file that fails the rules (for example a draft `0.4` file) earns 6. | Personal AI agents use this file to learn your APIs, agent and sign-in. |
| **llms.txt** | 12 | Markdown `/llms.txt` (8) with an H1 title (+2) and 3+ links (+2) | A curated map of your site for AI assistants. |
| **llms-full.txt** | 2 | Markdown `/llms-full.txt` | Full-text corpus for retrieval. |
| **AI crawler rules** (robots.txt) | 13 | Assistant and search agents allowed site-wide: OAI-SearchBot, ChatGPT-User, Claude-User, Claude-SearchBot, PerplexityBot, Perplexity-User, DuckAssistBot, MistralAI-User, Meta-ExternalFetcher (10, shared equally). Training crawlers: GPTBot, ClaudeBot, Google-Extended, Applebot-Extended, CCBot, Meta-ExternalAgent, Amazonbot, Bytespider (3). No robots.txt counts as allowed by default (10). Each bot's deciding rule and line number is reported. Meta documents that Meta-ExternalFetcher may bypass robots.txt for user-requested fetches, so its verdict shows your stated preference. | Blocked agents cannot read or cite you. |
| **Firewall lets AI agents in** | 6 | The homepage answers the documented ChatGPT-User user agent as it answers a browser | Many CDNs and WAFs silently block AI agents. |
| **schema.org JSON-LD** | 12 | Organization/LocalBusiness (6), WebSite (3), a page type such as Product, Service or FAQPage (3), in the raw homepage HTML | Lets agents understand who you are without guessing. |
| **Content readable without JavaScript** | 8 | 300+ visible words in the raw HTML (150+ earns 5) | Most AI crawlers do not run JavaScript. |
| **Markdown for agents** | 4 | Homepage returns Markdown for `Accept: text/markdown` | Agents read Markdown with far fewer tokens. |
| **XML sitemap** | 6 | `/sitemap.xml` or the `Sitemap:` line in robots.txt | Lets agents enumerate your pages. |
| **MCP discovery** | 8 | An MCP Server Card in `/.well-known/ai-catalog.json` (MCP Server Card extension, SEP-2127), a card at `/.well-known/mcp-server-card` or `/.well-known/mcp/server-card`, or an `mcp` API in poppy.json (8). `/.well-known/mcp.json` is a community convention outside the MCP spec (5). RFC 9728 metadata alone (2). | Agents can connect to your MCP server without manual setup. |
| **A2A Agent Card** | 3 | A valid Agent Card at `/.well-known/agent-card.json` per A2A 1.0 (`name`, `description`, `version`, `capabilities`, `defaultInputModes`, `defaultOutputModes`, `skills`, and `supportedInterfaces` with HTTPS `url`, `protocolBinding`, `protocolVersion`) earns 3. A valid pre-1.0 card (A2A 0.2/0.3: `protocolVersion` 0.x with a top-level `url`) earns 2; a card only at the old `/.well-known/agent.json` path earns 1; an A2A protocol listed in poppy.json earns 2. Fix priority is always low: it only matters if you run an agent. | Other AI agents can discover and call your agent. |
| **OpenAPI / API catalog** | 5 | OpenAPI/Swagger at `/openapi.json`, `/swagger.json`, `/openapi.yaml`, `/api/openapi.json` or `/v3/api-docs`, an RFC 9727 `/.well-known/api-catalog`, or an `openapi` API in poppy.json | Agents can call your API directly. |
| **Meta tags** | 4 | Title, meta description, canonical, and no `noindex`/`noai` robots directive (meta tag or X-Robots-Tag) | A homepage `noindex` drops you from search and AI search. |
| **HTTPS redirect + HSTS** | 2 | `http://` redirects to `https://` (same-site `http://` hops such as apex to www are followed, up to 3) and an HSTS header is sent | Basic trust signals. |
| **ai-plugin.json** (legacy) | 1 | The old ChatGPT plugin manifest | Lowest priority; superseded by poppy.json and MCP. |

#### Grade bands

| Grade | Score | Meaning |
|---|---|---|
| A | 80-100 | Agent-ready. Polish only. |
| B | 65-79 | Good foundation, a few gaps. |
| C | 50-64 | Readable by AI, but missing key agent signals. |
| D | 35-49 | Significant gaps; AI agents will struggle. |
| F | 0-34 | Largely invisible or unusable for AI agents. |

#### Fair scoring

A check only counts when the site actually answered: a real file, a 404, or a "soft 404" (an HTML page served with HTTP 200 for a missing file, which is never mistaken for llms.txt or JSON). Checks the site blocks (401/403/429, bot-challenge or JavaScript interstitial pages) or that time out are listed in `unverifiedChecks`, **excluded from the score and never reported as missing**. A domain with fewer than 9 of the 15 checks verified, or whose homepage cannot be read, is marked `unreachable` and **not charged**.

#### Not scored, but reported

- `aiUsagePreferences`: AI usage preferences the site declares, from IETF AIPREF `Content-Usage` rules in robots.txt and the `Content-Usage` HTTP header (Internet-Drafts, not yet RFCs), and `Content-Signal` lines (contentsignals.org). These are business choices, so they never change the score.
- `hiddenAgentInstructions`: where text aimed at AI agents is hidden in the homepage (location only, never copied).

### Output

**Dataset (one row per domain)**: `agentReadinessScore`, `grade`, `poppyJsonStatus`, `hasLlmsTxt`, `blockedAiCrawlers`, `firewallBlocksAiAgents`, `hasJsonLd`, `servesMarkdown`, `hasMcpDiscovery`, `hasA2aAgentCard`, `hasOpenApi`, `aiUsagePreferences`, `hasSitemap`, `isJsShell`, `rawHtmlWords`, `aiCrawlerRules` (per bot: verdict, group, deciding rule and line), `jsonLdTypes`, `hiddenAgentInstructions`, `topFix`, `fixes` (priority, points available, fix), `fixSnippets`, `isLead`, `pitchLine`, and `checks` (every check with status, points and evidence). Views: **Overview**, **Fixes & snippets**, **Details & evidence**, **Leads**, **Status & billing**.

**Key-value store**: `OUTPUT` (run summary: average and median score, grade distribution, pass rate per check, poppy.json adoption, most common fixes, ranking), `REPORT_HTML` (client-ready report, print to PDF), `REPORT_CSV`, `REPORT_MD`.

Example `checks.poppyJson` for a site serving a draft file:

```json
{
  "title": "poppy.json (Personal Agent Protocol 1.0)",
  "status": "invalid",
  "points": 6,
  "maxPoints": 14,
  "detail": "1 spec error(s): protocol_version is \"0.4\", a pre-1.0 draft that 1.x Personal Agents cannot use (spec 3.4). Upgrade the document to \"1.0\"",
  "evidence": { "url": "https://allbirds.com/.well-known/poppy.json", "httpStatus": 200, "contentType": "application/json", "snippet": "{\"protocol_version\":\"0.4\", ..." }
}
```

Example `fixSnippets.poppyJson` (minimal document that meets section 3.1: version, organization and the `web` channel):

```json
{
  "protocol_version": "1.0",
  "organization": { "name": "Example Domain", "domain": "example.com" },
  "web": {}
}
```

### Who uses it

- **SEO, GEO and AI-visibility agencies**: audit a whole client book or prospect list in one run, sort by score, and send the `pitchLine` or the HTML report.
- **Ecommerce teams**: check stores (including draft-0.4 `poppy.json` files) and fix JSON-LD, crawler and firewall gaps before agent shopping grows.
- **Developers and platform teams**: monitor `poppy.json`, MCP and OpenAPI discovery across many domains or sub-brands after each release.
- **Researchers**: measure adoption of llms.txt, poppy.json and MCP discovery across thousands of sites.

### Pricing

Pay per event: **US$0.02 per domain scored** (status `ok` or `partial`), which is US$20 per 1,000 domains. Apify subscribers pay less: Bronze US$0.018, Silver US$0.016, Gold and above US$0.015. Invalid, duplicate, unreachable, timed-out, DNS-error and private-address rows are **always free**. A tiny standard Actor-start fee applies. Set **Maximum cost per run** in the run options to cap spend.

| Domains | Cost (free plan) |
|---|---|
| 10 | US$0.20 |
| 100 | US$2.00 |
| 1,000 | US$20.00 |

### Use it from the API, n8n, Make, Zapier or an AI agent

```bash
curl -X POST "https://api.apify.com/v2/acts/ausbizstack~agent-readiness-bulk-auditor/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"domains": ["example.com", "yourclient.com"]}'
```

```js
import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('ausbizstack/agent-readiness-bulk-auditor').call({ domains: ['example.com'] });
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items.map((i) => [i.domain, i.agentReadinessScore, i.grade]));
```

```python
from apify_client import ApifyClient
client = ApifyClient("<APIFY_TOKEN>")
run = client.actor("ausbizstack/agent-readiness-bulk-auditor").call(run_input={"domains": ["example.com"]})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["domain"], item["agentReadinessScore"], item["grade"])
```

- **AI agents (MCP)**: add `https://mcp.apify.com/?actors=ausbizstack/agent-readiness-bulk-auditor` to Claude, Cursor or any MCP client.
- **No-code**: use the Apify app in n8n, Make or Zapier, or schedule this Actor in Apify and send results to Google Sheets, Slack or a webhook.

### FAQ

**Does it crawl the whole site?** No. It checks the homepage and about 22 well-known discovery URLs per domain, which keeps it fast and cheap for big lists.

**How big a list can I run?** Up to 10,000 domains per run. Inputs are normalised and de-duplicated before any work. On the default 256 MB, up to 11 domains run in parallel; more memory allows up to 50. If the run's time limit or your maximum cost is reached, the remaining domains get free `timeout` rows, so the dataset always has one row per input.

**Why is a check "unverified"?** The site blocked or timed out on that URL. It is excluded from the score rather than counted as missing, so blocked sites are not unfairly marked down.

**Is the firewall test definitive?** It sends ChatGPT-User's documented user agent from an Apify server. Firewalls that admit only verified agent IP ranges will also show as blocking here, so treat a failure as "check your bot rules", not proof.

**Is Personal Agent Protocol final?** Version 1.0 was published on 14 November 2026. Sierra's announcement is dated 13 November 2026. This Actor validates the 1.0 discovery rules and flags older draft versions as not 1.0-compatible.

**Does it change or log in to anything?** No. It sends only public, unauthenticated GET requests. It never signs in, never calls an MCP server, agent or API, and never follows URLs found inside poppy.json or an Agent Card. IP addresses, localhost and internal names are refused at input. Every connection, including each redirect hop, is checked: anything that points at a private, loopback, link-local, carrier-grade NAT, reserved or cloud-metadata address is refused, including host names whose DNS resolves there. Responses are size-capped (1.5 MB for the homepage, less for other files) and time-limited.

**What are "hidden agent instructions"?** Text aimed at AI agents hidden in the homepage (hidden elements, HTML comments, zero-width characters). It is reported as a location only and never copied. It is informational and does not affect the score.

**Can I white-label the report?** Yes. Set `reportTitle`; the HTML report has no AusBizStack branding in the body.

### Sources

- Personal Agent Protocol 1.0 spec: https://personalagentprotocol.org/docs/1.0/spec (changelog: https://personalagentprotocol.org/docs/changelog.md); announcement: https://sierra.ai/blog/releasing-poppy
- llms.txt proposal: https://llmstxt.org/
- MCP Server Cards and AI Catalog discovery (SEP-2127): https://github.com/modelcontextprotocol/modelcontextprotocol/blob/main/seps/2127-mcp-server-cards.md
- A2A protocol 1.0.0 (AgentCard, section 4.4.1): https://a2a-protocol.org/latest/specification/ · well-known discovery: https://a2a-protocol.org/latest/topics/agent-discovery/
- IETF AIPREF Content-Usage (Internet-Draft): https://datatracker.ietf.org/doc/draft-ietf-aipref-attach/ and vocabulary https://datatracker.ietf.org/doc/draft-ietf-aipref-vocab/ · Content Signals: https://contentsignals.org/
- OAuth 2.0 Protected Resource Metadata (RFC 9728): https://www.rfc-editor.org/rfc/rfc9728 · API Catalog (RFC 9727): https://www.rfc-editor.org/rfc/rfc9727 · robots.txt (RFC 9309): https://www.rfc-editor.org/rfc/rfc9309

The score reflects publicly visible technical signals and is a guide, not a guarantee of AI search placement.

Built by AusBizStack.

# Actor input Schema

## `domains` (type: `array`):

One per line (pasting a comma- or space-separated list into one line also works). Bare domains (acme.com), full URLs (https://www.acme.com/pricing) and email addresses (hello@acme.com) all work: paths, ports, query strings, case and 'www.' are normalised and duplicates are removed. Public websites only: IP addresses, localhost and internal names are refused. Duplicate and invalid lines get a free row. Non-ASCII (IDN) domains are supported.

## `includeFixSnippets` (type: `boolean`):

Adds a draft llms.txt (built from the site's own homepage links), a minimal valid poppy.json, robots.txt lines for blocked AI agents and an Organization JSON-LD template to each row, only where the check failed.

## `leadThreshold` (type: `integer`):

Domains scoring below this get isLead = true plus a one-line pitchLine summarising their biggest gaps. Useful for agencies building prospect lists. Set 0 to disable.

## `reportTitle` (type: `string`):

Heading for the shareable HTML/Markdown reports saved in the key-value store (REPORT_HTML, REPORT_MD). Use your own brand or the client's name.

## `maxConcurrency` (type: `integer`):

How many domains are audited at the same time (1-50). Each domain makes about 23 small GET requests, at most 8 at a time per site. On 256 MB of memory this is capped at 11; give the run more memory to go faster.

## `requestTimeoutSecs` (type: `integer`):

Timeout for each HTTP request.

## `perDomainTimeoutSecs` (type: `integer`):

Hard cap per domain. When it is hit the row is still saved: unfinished checks are marked unverified and excluded from the score (never reported as missing).

## Actor input object example

```json
{
  "domains": [
    "example.com"
  ],
  "includeFixSnippets": true,
  "leadThreshold": 50,
  "reportTitle": "AI agent-readiness report",
  "maxConcurrency": 10,
  "requestTimeoutSecs": 5,
  "perDomainTimeoutSecs": 10
}
```

# Actor output Schema

## `overview` (type: `string`):

No description

## `fixes` (type: `string`):

No description

## `summary` (type: `string`):

No description

## `reportHtml` (type: `string`):

No description

## `reportCsv` (type: `string`):

No description

## `leads` (type: `string`):

No description

## `reportMd` (type: `string`):

No description

# API

You can run this Actor programmatically using our API. Below are code examples in JavaScript, Python, and CLI, as well as the OpenAPI specification and MCP server setup.

## JavaScript example

```javascript
import { ApifyClient } from 'apify-client';

// Initialize the ApifyClient with your Apify API token
// Replace the '<YOUR_API_TOKEN>' with your token
const client = new ApifyClient({
    token: '<YOUR_API_TOKEN>',
});

// Prepare Actor input
const input = {
    "domains": [
        "example.com"
    ]
};

// Run the Actor and wait for it to finish
const run = await client.actor("ausbizstack/agent-readiness-bulk-auditor").call(input);

// Fetch and print Actor results from the run's dataset (if any)
console.log('Results from dataset');
console.log(`💾 Check your data here: https://console.apify.com/storage/datasets/${run.defaultDatasetId}`);
const { items } = await client.dataset(run.defaultDatasetId).listItems();
items.forEach((item) => {
    console.dir(item);
});

// 📚 Want to learn more 📖? Go to → https://docs.apify.com/api/client/js/docs

```

## Python example

```python
from apify_client import ApifyClient

# Initialize the ApifyClient with your Apify API token
# Replace '<YOUR_API_TOKEN>' with your token.
client = ApifyClient("<YOUR_API_TOKEN>")

# Prepare the Actor input
run_input = { "domains": ["example.com"] }

# Run the Actor and wait for it to finish
run = client.actor("ausbizstack/agent-readiness-bulk-auditor").call(run_input=run_input)

# Fetch and print Actor results from the run's dataset (if there are any)
print(f"💾 Check your data here: https://console.apify.com/storage/datasets/{run.default_dataset_id}")
for item in client.dataset(run.default_dataset_id).iterate_items():
    print(item)

# 📚 Want to learn more 📖? Go to → https://docs.apify.com/api/client/python/docs/quick-start

```

## CLI example

```bash
echo '{
  "domains": [
    "example.com"
  ]
}' |
apify call ausbizstack/agent-readiness-bulk-auditor --silent --output-dataset

```

## MCP server setup

```json
{
    "mcpServers": {
        "apify": {
            "type": "http",
            "url": "https://mcp.apify.com/?tools=fetch-actor-details,ausbizstack/agent-readiness-bulk-auditor"
        }
    }
}
```

The hosted server signs you in with OAuth on first connect, so no API token belongs in this config. Clients without OAuth support can send an `Authorization: Bearer <APIFY_API_TOKEN>` header instead, using a token from API & Integrations in Apify Console (https://console.apify.com/settings/integrations).

## OpenAPI specification

Download the OpenAPI definition: https://api.apify.com/v2/actors/m0ggdfYePyepcwJhL/builds/wNSuMdPhu2lO89vmW/openapi.json
