Agent Data Tools: Web, documents and search

by io.github.Nutshell2000

Static web page to text, Markdown, code and tables

extract

Read public static HTML, Markdown or plain text. HTML includes code, tables, links, heading IDs, image alt text and unverified author/date provenance. Optional exact HTML section ID. Default 20,000 output characters, maximum 100,000; input 2 MiB. No JavaScript rendering, authentication or paywall access. Empty pages, invalid sections, browser challenges and execution errors are uncharged. Use to read docs or articles into context. Example: url=https://example.com, maxChars=20000, format=text, section=retry-policy. Price $0.002/call via x402 USDC or a card key (/credits), first 25 free; failed calls free. Reads or computes only; not marked read-only because each call is charged.

Input Schema

TableJSON Schema

Name	Required	Description
`url`	Yes	Absolute http(s) URL of the page to fetch
`format`	No	Returned representation: text (default), markdown, or both. HTML structured details are included in every format; Markdown/plain sources have empty details.
`section`	No	Optional exact visible HTML ID within primary content, 1-500 characters. Heading scope includes subsections until next same/higher-level heading; section/article/main/div scope is its subtree. Missing/ambiguous/unsupported targets fail uncharged. URL fragments alone do not select a section.
`maxChars`	No	Maximum characters of body text to return (500-100000, default 20000)

Output Schema

TableJSON Schema

Name	Required	Description
`url`	Yes
`lang`	Yes
`text`	No
`links`	Yes
`title`	Yes
`format`	Yes
`status`	Yes
`tables`	Yes
`section`	Yes	Null for full primary-content reading. Otherwise the explicitly requested scope; title/description/canonical remain page-level metadata. Heading scope ends before the next same/higher-level heading, including nested subsections; container scope is its subtree.
`finalUrl`	Yes
`headings`	Yes
`markdown`	No
`warnings`	Yes
`canonical`	Yes
`fetchedAt`	Yes
`truncated`	Yes	Input or requested primary representation is incomplete. Structured detail limits are reported separately in warnings.
`codeBlocks`	Yes
`hashFormat`	Yes
`textLength`	Yes	Full extracted text length before the character cap, within the input limit.
`contentHash`	Yes
`contentType`	Yes
`description`	Yes
`linkBaseUrl`	Yes	Effective source-derived HTTP(S) base for links: the first supported HTML base href outside inert/foreign subtrees, otherwise finalUrl. Credentials and non-HTTP(S) base values are ignored with a warning. HTML links/canonical use this base; raw Markdown keeps relative syntax and can be resolved using this field. No linked resource is fetched. This static interpretation does not evaluate scripts or Content Security Policy.
`sourceFormat`	Yes	Source interpretation. Markdown is identified by text/markdown, or a .md/.markdown final path served as text/plain (warned). Markdown syntax is retained verbatim within the cap, not converted to HTML or parsed for structured details; text also retains that syntax. HTML section IDs are unsupported for Markdown/plain text.
`sourceMetadata`	Yes	Page-level source declarations, not verified authorship or freshness. Head author/Open Graph date tags and bounded direct Schema.org JSON-LD Article/WebPage-like nodes only; exact-URL matches preferred, otherwise only one identity-free candidate. Author @id references resolve only to unambiguous local Person/Organization names, with both paths recorded. No external context/reference fetching or general JSON-LD expansion. Separate 2000-character value+provenance budget; 5 authors/10 dates. Original date strings are not normalized. Empty for Markdown/plain sources; unrelated entities and ambiguous nodes are omitted with warnings.
`contentSelection`	Yes

Schema Changelog

Changes observed during successful MCP inspections.

First observedOct 19, 2026

TDQS

A4.6/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Goes well beyond the annotations: it discloses the 20k/100k output ceiling, 2 MiB input cap, uncharged failure classes (empty, invalid section, browser challenge, execution error), pricing model, and free-tier behavior. It also explicitly reconciles the surprising readOnlyHint=false by explaining that each call is charged.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Front-loaded with purpose, then limits, exclusions, usage, example and price in a tight sequence with no filler sentences. It is dense and long for a single description, but each block carries operational information an agent needs before calling.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

An output schema exists so return values need no prose, and the description still covers scope, limits, failure semantics, cost and the annotation rationale. Nothing required to invoke the tool correctly is missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, which sets a baseline of 3, but the description adds the 2 MiB input limit (absent from the schema), restates the default/max char budget in prose, and gives a worked example binding url, maxChars, format and section together. Format and section semantics are largely left to the schema.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb (read/extract) and resource (public static HTML, Markdown, plain text) plus the exact payload returned (code, tables, links, heading IDs, alt text, provenance). The static-only scope implicitly separates it from the rendered sibling, so an agent can route correctly without opening a schema.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

"Use to read docs or articles into context" gives a concrete use case, and "No JavaScript rendering, authentication or paywall access" tells the agent when this tool will fail and an alternative (extract_rendered) is needed. It stops short of naming the alternative explicitly, so it is not fully prescriptive.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All Agent Data Tools: Web, documents and search tools Inspect

Try in Browser

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.