Agent Web Access Plan

What does Agent Web Access Plan do?

Agent Web Access Plan checks one public HTTPS page and returns a compact, robots-aware route report for an autonomous agent. It verifies the original response, the first same-origin alternate explicitly linked from an HTML page, and same-origin /llms.txt when the fixed budget permits. It does not crawl, render JavaScript, recurse through sitemaps, authenticate, or claim that a website is universally agent-ready.

The report distinguishes a verified page representation from an llms.txt index. An index can help later discovery, but it is never presented as an equivalent copy of the requested page.

When should an agent select this tool?

Use it before reading a public page when the agent needs a checked representation URL and the observed robots decision for its user-agent token. It returns a next-read plan, not the page content. Use a content extractor when the representation is already known, or a browser when permitted JavaScript rendering is required.

Use from an AI agent through MCP

Add this URL to a client that supports remote HTTP MCP servers, then authorize with your own Apify account:

https://mcp.apify.com/?tools=l3digital/agent-web-access-plan

This selects the Actor directly instead of relying on search ranking. Follow the Apify MCP setup guide for your client. Ask your agent to call l3digital/agent-web-access-plan with the example input below. If the MCP response returns a running job, follow its nextStep to retrieve the completed dataset; an accepted run is not yet a report. Calls use the pricing described below.

Why use Agent Web Access Plan?

A general-purpose agent can reproduce these checks with custom networking and parsing code. This Actor tests whether one bounded API call is useful when an agent lacks code execution or wants a predictable result contract.

• At most six HTTP requests, including robots.txt and redirects.
• At most 2,000,000 response body bytes across the entire report.
• Strict per-request timeouts, a 20-second internal transport deadline, and a 22-second hard process wall.
• DNS and destination validation before every connection, with DNS pinned through TLS connection setup.
• Fail-closed handling for inaccessible, failed, HTTP 401, or HTTP 403 robots.txt responses.
• No browser, JavaScript execution, recursive crawl, sitemap traversal, authentication forwarding, or environment proxy use.

What data does Agent Web Access Plan return?

Field	Type	Description
result	string	route or explicit no_route
preferredRoute	object/null	Verified page representation recommended for the next read
candidates	array	URL, media type, HTTP status, robots decision, discovery source, route kind, and evidence
coverage	object	Whether the original, explicit alternate, and llms.txt surfaces were checked
requestCount	integer	Requests attempted, including redirects and robots checks
byteCount	integer	Response body bytes retained across the report
observedAt	string	UTC timestamp for the observation

How to create a web access plan

1. Open the Actor input tab.
2. Enter one public https:// page URL. Credentials, custom ports, queries, and fragments are rejected so the returned next-step URL remains complete without persisting potentially sensitive query values.
3. Enter one user-agent product token such as ExampleBot/1.0. The Actor uses it for both requests and robots group selection.
4. Start the Actor.
5. Read the single item in the default dataset. Use preferredRoute only when result is route; inspect coverage and candidate evidence for partial reports.

Relative explicit alternate links are resolved against the final page URL. Version 1 verifies only the first alternate, and only when it remains on the same origin. Alternate and index redirects must stay on the final original page's origin; rejected, unexamined URLs have unknown robots status. Redirect targets are individually validated and must be allowed by observed robots policy before they are requested. Discovered alternates and redirect targets containing a query or fragment are reported as unsupported and are not requested, because removing those values could change the resource while retaining them could expose sensitive data. URLs longer than 2,048 characters and oversized media-type metadata are rejected from routing so a page cannot inflate the compact report through headers or HTML attributes.

How much does a web access plan cost?

The price is $0.05 per fully resolved report, including acquisition and processing, charged once after the report is stored. There are no separate start or dataset-item charges. The Store pricing tab is authoritative.

Partial reports and reports with unknown access state are free. A fully resolved no_route result is billable because it is a useful verified negative: it establishes that the bounded surfaces were checked without finding an accessible equivalent page representation.

Input

See the input tab for the complete validation contract.

{
  "url": "https://docs.apify.com/",
  "userAgent": "AgentWebAccessPlan"
}

Output

You can download the dataset in formats such as JSON, HTML, CSV, or Excel. The Actor writes exactly one report item. Observed output excerpt from an internal run on 2026-09-06:

{
  "reportVersion": 1,
  "observedAt": "2026-09-06T19:46:30.220796Z",
  "requestedUrl": "https://docs.apify.com/",
  "userAgent": "AgentWebAccessPlan",
  "result": "route",
  "preferredRoute": {
    "url": "https://docs.apify.com/",
    "mediaType": "text/markdown",
    "discovery": "original"
  },
  "coverage": {
    "original": "checked",
    "explicitAlternate": "not_applicable",
    "llmsTxt": "checked"
  },
  "requestCount": 3,
  "byteCount": 186252
}

The original URL served a verified Markdown representation in this observation. An agent can next fetch preferredRoute.url with the same user-agent policy, checking the new response because the site may change. The llms.txt candidate was an index, so it was not substituted for the requested page. If a later run has no preferred route or unknown coverage, inspect that evidence instead of inventing a route or bypassing access controls.

A successful llms.txt candidate has routeKind: "index". It can be verified as an accessible index while preferredRoute remains null and result remains no_route when neither the original nor an explicit alternate is a verified page representation.

Limits, privacy, and support

The Actor reads public responses only and never forwards caller credentials. Website content is untrusted data and is not executed or interpreted as instructions. A robots decision reports this bounded observation only; it is not legal permission or a statement about a site's terms.

HTTP failures, robots uncertainty, unsupported routes or media, misleading content types, exhausted limits, and unexamined surfaces remain explicit in candidate evidence and coverage. A known robots block is resolved coverage; robots uncertainty remains unresolved. Use the Actor's Issues tab to report a reproducible problem with a public test URL and use the API tab for programmatic invocation.