biorxiv-mcp-server

Get Preprint Full Text

biorxiv_get_fulltext

Read-only

Retrieve a preprint's full text as best-effort Markdown, extracted from its rendered HTML article page. Resolves the latest version via the details API, then fetches and extracts the body — abstract, sections, and references. bioRxiv and medRxiv share the 10.1101/ DOI prefix, so server="both" (the default) resolves the DOI against both in parallel and the response reports which server answered. This is HTML-to-Markdown extraction, not structured JATS: section structure is approximate and not guaranteed. Long articles exceed a single response, so use offset and limit to page through them (the response reports totalChars, remainingChars, and hasMore); paging is cheap because the extracted article is cached per version for an hour after the first read, so only the first chunk pays for a fetch. Not every preprint has an extractable HTML page — some are PDF-only and some origins block programmatic access — in which case a fulltext_unavailable error routes you to biorxiv_get_preprint for the title, abstract, and metadata. For a preprint that has been published in a journal, the journal's version may have richer full text elsewhere.

Input Schema

TableJSON Schema

Name	Required	Description	Default
`doi`	Yes	Preprint DOI (e.g. 10.1101/2024.05.28.596311 or 10.64898/2026.05.07.723463). The latest version is resolved automatically.
`limit`	No	Maximum number of characters to return in this chunk. Default 20,000; increase toward 50,000 for large context windows. Check the length field for the actual count returned.
`offset`	No	Character offset into the full extracted text at which to start reading. 0 returns the beginning. To read the next chunk, use offset = prior_offset + prior_length (the length field from the previous response).
`server`	No	Server the preprint was posted on. "both" (default) checks bioRxiv and medRxiv in parallel to resolve the DOI — the full-text fetch itself only ever targets whichever server resolved, and the output server field names it.	both

Output Schema

TableJSON Schema

Name	Required	Description
`cap`	No	The limit (max characters) applied to this chunk.
`doi`	No	The resolved preprint DOI.
`error`	No	Present when the call failed. Absent on success.
`shown`	No	Characters returned in this chunk.
`title`	No	Article title detected during extraction. Absent when the page exposed none.
`length`	No	Number of characters returned in this chunk.
`notice`	No	Paging guidance when the content was truncated — how to fetch the next chunk.
`offset`	No	Character offset into the full extracted text where this chunk begins.
`server`	No	Server the preprint was resolved on.
`content`	No	The requested chunk of full text as best-effort Markdown extracted from the rendered HTML page. Section structure is approximate — this is not JATS.
`hasMore`	No	True when more text follows this chunk. When true, call again with offset = offset + length.
`version`	No	Preprint version whose full text was retrieved (the latest revision).
`sourceUrl`	No	The full-text HTML page the content was extracted from.
`truncated`	No	True when this chunk was capped by limit and more text remains.
`wordCount`	No	Approximate word count of the FULL extracted article as reported by the extractor (not just the returned chunk). Absent when the extractor reported none.
`totalChars`	No	Total characters in the full extracted text. Use with offset and length to page through long articles.
`contentFormat`	No	How content was produced: Markdown extracted from the rendered HTML article page (constant).
`remainingChars`	No	Characters remaining after this chunk (totalChars - offset - length). 0 means this chunk reaches the end.

Schema Changelog

Changes observed during successful MCP inspections.

First observedSep 19, 2026

TDQS

A4.9/5.0

Behavior5/5

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

Beyond the readOnlyHint and openWorldHint annotations, the description discloses best-effort Markdown quality, approximate section structure, per-version caching for an hour, parallel server resolution, error behavior, and fallback conditions. This is rich behavioral context that an agent cannot infer from the annotations or schema.

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

Conciseness5/5

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

The description is long but every sentence carries distinct information: extraction method, server resolution, quality caveat, paging/caching, error fallback, and journal alternative. It is front-loaded with the core retrieval purpose and progresses logically through operational details.

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?

For a tool with this complexity, the description covers the key operational facts: DOI resolution, paging through long articles, likely error modes and their fallback, and the approximate nature of the extracted structure. Since an output schema exists, detailed return-value documentation is not needed.

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?

With 100% schema description coverage, the baseline is 3, and the description adds value by explaining the paging protocol (offset = prior_offset + prior_length), the caching implications for later chunks, and how server resolution works. It complements the schema rather than merely restating it.

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?

The first sentence states a specific verb+resource: 'Retrieve a preprint's full text as best-effort Markdown, extracted from its rendered HTML article page,' and then names the concrete content extracted (abstract, sections, references). It differentiates itself from siblings by explicitly contrasting with biorxiv_get_preprint and the journal-published version, and by stating this is not structured JATS.

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

Usage Guidelines5/5

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

The description gives explicit routing guidance: when a fulltext_unavailable error occurs, use biorxiv_get_preprint for metadata; when a preprint has a journal version, richer full text may live elsewhere. It also specifies when to use offset and limit for paging rather than one call, making the selection and invocation conditions clear.

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

All biorxiv-mcp-server 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.