The Unpaywall MCP Server enables AI clients to search, retrieve metadata, and access full-text content for academic papers using the Unpaywall and OpenAlex APIs.
Fetch metadata by DOI (unpaywall_get_by_doi): Retrieve full Unpaywall metadata (title, OA status, locations, etc.) for any article using its DOI, DOI URL, or doi: prefix.
Search article titles (unpaywall_search_titles): Search for academic articles by title query, with optional filtering by open-access or closed-access status, and paginated results (50 per page).
Get open-access fulltext links (unpaywall_get_fulltext_links): For a given DOI, retrieve the best available OA PDF URL and landing page URL, along with all OA location metadata.
Download and extract PDF text (unpaywall_fetch_pdf_text): Download and extract text from the best OA PDF for a DOI or a directly provided PDF URL, with configurable truncation (default 20,000 characters).
Provides tools for fetching academic paper metadata by DOI, searching article titles, retrieving open access fulltext links, and downloading/extracting text from open access PDFs through the Unpaywall database.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Unpaywall MCP Server search for papers about quantum computing and machine learning"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
An MCP (Model Context Protocol) server exposing Unpaywall tools so AI clients can:
Fetch metadata by DOI
Search article titles
Retrieve best OA fulltext links
Download and extract text from OA PDFs
Add this to your MCP client config (Claude Desktop example):
{
"mcpServers": {
"unpaywall": {
"command": "npx",
"args": ["-y", "unpaywall-mcp"],
"env": { "UNPAYWALL_EMAIL": "you@example.com" }
}
}
}Then try the tools: unpaywall_search_titles, unpaywall_get_fulltext_links, unpaywall_fetch_pdf_text.
You don't need to clone this repo or run npm install — npx handles fetching and caching on first call.
Related MCP server: ScholarScope MCP
Node.js 18+ (for npx)
An email address for Unpaywall / OpenAlex requests (required by Unpaywall, used for the OpenAlex polite pool).
End users should use the npx config above. Contributors building from source:
npm install
npm run build
UNPAYWALL_EMAIL=you@example.com npm start # stdio transport, as required by MCP clientsHot-run (no build step):
UNPAYWALL_EMAIL=you@example.com npm run devDescription: Fetch Unpaywall metadata for a DOI
Input schema:
doi (string, required): e.g. 10.1038/nphys1170
email (string, optional): overrides UNPAYWALL_EMAIL if provided
Output: JSON response from Unpaywall
Description: Search article titles and return Unpaywall-style OA metadata for each hit (50 results/page)
Input schema:
query (string, required): title query
is_oa (boolean, optional): if true, only OA results; if false, only closed; omit for all
page (integer >= 1, optional): page number
email (string, optional): overrides UNPAYWALL_EMAIL
Output: JSON matching the Unpaywall search shape — results[].response is a DOI-style record (doi, title, is_oa, oa_status, best_oa_location, oa_locations), with score and snippet per result. _source: "openalex" marks the upstream.
Note: Backed by OpenAlex's /works endpoint because Unpaywall's own /v2/search has been returning HTTP 500 since its May 2025 rewrite. Unpaywall now runs as a subroutine of OpenAlex, so this is the canonical modern equivalent.
Description: Return the best OA PDF URL and Open URL for a DOI, plus all OA locations
Input schema:
doi (string, required)
email (string, optional): overrides UNPAYWALL_EMAIL
Output: JSON with fields: best_pdf_url, best_open_url, best_oa_location, oa_locations, and select metadata
Description: Download and extract text from the best OA PDF for a DOI, or from a provided pdf_url
Input schema:
pdf_url (string, optional): direct PDF URL (takes precedence)
doi (string, optional): used to resolve best OA PDF if pdf_url not provided
email (string, optional): required if using doi and no UNPAYWALL_EMAIL env var
truncate_chars (integer >= 1000, optional): max characters of extracted text to return (default 20000)
Output: JSON with text (possibly truncated), length_chars, truncated, pdf_url, and PDF metadata
When using this server from an MCP-enabled LLM client, ask the model to:
Search then fetch: Use unpaywall_search_titles with a concise title phrase; select a result; then call unpaywall_get_fulltext_links or unpaywall_fetch_pdf_text on the chosen DOI.
Prefer OA: Pass is_oa: true in searches when you only want open-access.
Control size: Set truncate_chars in unpaywall_fetch_pdf_text (default 20000) and summarize long texts before proceeding.
Be resilient: If the best PDF URL is missing, fall back to best_open_url and extract content from the landing page (outside this server).
Respect rate limits: Space requests if making many calls; reuse earlier responses instead of repeating the same call.
Good user instructions to the LLM:
"Find 3 OA papers about 'foundation models in biomedicine', then extract and summarize the introduction of the best one."
"Search for 'Graph Neural Networks survey 2024', filter to OA if possible, then fetch the PDF text and produce a 10-bullet summary."
Depending on your MCP client, the structure differs; the core payloads are:
// Search titles
{
"name": "unpaywall_search_titles",
"arguments": {
"query": "graph neural networks survey",
"is_oa": true,
"page": 1
}
}// Get best OA links for a DOI
{
"name": "unpaywall_get_fulltext_links",
"arguments": {
"doi": "10.48550/arXiv.1812.08434"
}
}// Fetch and extract PDF text (by DOI)
{
"name": "unpaywall_fetch_pdf_text",
"arguments": {
"doi": "10.48550/arXiv.1812.08434",
"truncate_chars": 20000
}
}Recommended (no-build) config for Claude Desktop using npm/npx:
{
"mcpServers": {
"unpaywall": {
"command": "npx",
"args": ["-y", "unpaywall-mcp"],
"env": {
"UNPAYWALL_EMAIL": "you@example.com"
}
}
}
}Alternative (local repo) config using the compiled dist:
{
"mcpServers": {
"unpaywall": {
"command": "node",
"args": ["/absolute/path/to/dist/index.js"],
"env": {
"UNPAYWALL_EMAIL": "you@example.com"
}
}
}
}After adding, ask your client to list tools and try:
unpaywall_search_titles with a query
unpaywall_get_fulltext_links with a doi
unpaywall_fetch_pdf_text with a doi (or pdf_url)
Respect Unpaywall's rate limits and usage guidelines: https://unpaywall.org/products/api
The server uses stdio transport and @modelcontextprotocol/sdk.
Set UNPAYWALL_EMAIL or pass email per call so Unpaywall can contact you about usage.
# 1) Build the project (also runs automatically on publish)
npm run build
# 2) Bump version (choose patch/minor/major)
npm version patch
# 3) Publish (ensure you are logged in: npm login)
npm publish --access public
# 4) Tag a release on GitHub (optional, recommended)Users can then configure their MCP client with npx -y unpaywall-mcp as shown above. No clone or build required.
Download and extract text from best OA PDF for a DOI, or from a provided PDF URL.
| Name | Required | Description | Default |
|---|---|---|---|
| doi | No | DOI string or DOI URL. Used if pdf_url is not provided. | |
| pdf_url | No | Direct PDF URL to download and parse (takes precedence over DOI). | |
| No | Email to identify requests to Unpaywall (required when resolving via DOI). | ||
| truncate_chars | No | Max characters of extracted text to return (default 20000). |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It mentions key behaviors like downloading, extracting text, and precedence rules, but lacks details on error handling, rate limits, authentication needs (beyond the email parameter), or what 'best OA PDF' means. This provides basic context but is incomplete for a tool with potential external dependencies.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, efficient sentence that front-loads the core purpose ('Download and extract text') and includes essential qualifiers. Every word earns its place, with no wasted text, making it highly concise and well-structured.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given no annotations and no output schema, the description provides basic purpose and parameter context but lacks completeness. It doesn't explain return values (e.g., text format, error responses) or behavioral details like rate limits, which are important for a tool interacting with external services. This is adequate but has clear gaps for effective agent use.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so the schema fully documents all parameters. The description adds no additional semantic meaning beyond what's in the schema (e.g., it doesn't explain format details or usage nuances). This meets the baseline for high schema coverage but doesn't enhance parameter understanding.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the action ('Download and extract text') and the resource ('best OA PDF for a DOI, or from a provided PDF URL'), making the purpose understandable. However, it doesn't explicitly differentiate from sibling tools like 'unpaywall_get_fulltext_links' which might provide links rather than extracted text, so it misses full sibling distinction.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage by mentioning alternatives ('DOI, or from a provided PDF URL') and precedence rules ('pdf_url takes precedence over DOI'), but it doesn't explicitly state when to use this tool vs. siblings like 'unpaywall_get_by_doi' or provide clear exclusions. This leaves some ambiguity in tool selection.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Fetch Unpaywall metadata for a DOI (accepts DOI, DOI URL, or 'doi:' prefix). Requires an email address via env UNPAYWALL_EMAIL or the optional 'email' argument.
| Name | Required | Description | Default |
|---|---|---|---|
| doi | Yes | DOI string or DOI URL, e.g. 10.1038/nphys1170 or https://doi.org/10.1038/nphys1170 | |
| No | Email to identify your requests to Unpaywall (optional override) |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden. It discloses key behavioral traits: the tool requires an email address (via env or argument) for identification to Unpaywall, implying authentication needs. However, it lacks details on rate limits, error handling, or response format, leaving gaps in behavioral context.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is appropriately sized and front-loaded, with two sentences that efficiently convey purpose, input formats, and requirements without any wasted words. Every sentence adds essential information, making it highly concise and well-structured.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's moderate complexity (2 parameters, no output schema, no annotations), the description is adequate but incomplete. It covers authentication needs and input formats but lacks details on output (metadata structure), error cases, or performance constraints, which would be helpful for an AI agent.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so the schema already documents both parameters thoroughly. The description adds minimal value beyond the schema by mentioning the 'doi:' prefix as an accepted format and noting the email requirement, but does not provide additional syntax or usage details for parameters.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the specific action ('Fetch Unpaywall metadata') and resource ('for a DOI'), distinguishing it from sibling tools like fetching PDF text or searching titles. It specifies the exact input formats accepted (DOI, DOI URL, or 'doi:' prefix), making the purpose unambiguous.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides clear context on when to use this tool (for fetching metadata by DOI) and mentions an alternative (using the 'email' argument vs. env variable), but does not explicitly state when not to use it or differentiate from siblings like 'unpaywall_get_fulltext_links' beyond the metadata focus.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Given a DOI, return best open-access links (best PDF URL and open URL) plus Unpaywall locations metadata.
| Name | Required | Description | Default |
|---|---|---|---|
| doi | Yes | DOI string or DOI URL | |
| No | Email to identify your requests to Unpaywall (optional override) |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It describes what the tool returns (links and metadata) but does not mention rate limits, authentication needs (though 'email' is optional), error handling, or data freshness. It adds basic context but lacks details on operational traits like performance or constraints.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single, efficient sentence that front-loads the core functionality ('return best open-access links') and includes key details (PDF URL, open URL, metadata) without waste. Every word earns its place, making it highly concise and well-structured.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's moderate complexity (2 parameters, no output schema, no annotations), the description is reasonably complete. It specifies the input (DOI) and output (links and metadata), but without an output schema, it could benefit from more detail on return values (e.g., structure or examples). It covers the essentials but has minor gaps in output clarification.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so the schema already documents both parameters ('doi' as required string and 'email' as optional override). The description adds no additional meaning beyond what the schema provides, such as format examples or usage tips for parameters. Baseline 3 is appropriate when schema does the heavy lifting.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the specific action ('return best open-access links') and resource ('given a DOI'), distinguishing it from siblings like 'unpaywall_fetch_pdf_text' (which extracts text) and 'unpaywall_search_titles' (which searches titles). It explicitly mentions the output includes 'best PDF URL and open URL' plus metadata, making the purpose highly specific and differentiated.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage context by specifying 'given a DOI' for retrieving open-access links, but it does not explicitly state when to use this tool versus alternatives like 'unpaywall_get_by_doi' (which might return different data) or 'unpaywall_fetch_pdf_text' (which focuses on text extraction). It provides clear context but lacks explicit exclusions or named alternatives.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Search Unpaywall for article titles matching a query. Supports optional is_oa filter and pagination (50 results per page).
| Name | Required | Description | Default |
|---|---|---|---|
| query | Yes | Title search query (supports phrase, boolean operators per Unpaywall docs) | |
| is_oa | No | If true, only return OA results; if false, only closed; omit for all | |
| page | No | Page number (50 results per page) | |
| No | Email to identify your requests to Unpaywall (optional override) |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries the full burden of behavioral disclosure. It mentions 'Supports optional is_oa filter and pagination (50 results per page)', which provides useful operational context about filtering capabilities and pagination behavior. However, it doesn't address important behavioral aspects like rate limits, authentication requirements (beyond the optional email parameter), error conditions, or what the search results actually contain.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is extremely concise (two sentences) and front-loaded with the core purpose. Every sentence earns its place: the first establishes what the tool does, and the second provides key operational context about filters and pagination. There's zero waste or redundancy.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a search tool with 4 parameters, 100% schema coverage, and no output schema, the description provides adequate but minimal context. It covers the basic purpose and mentions key operational features (filtering, pagination), but doesn't explain what the search results contain, how comprehensive the search is, or any limitations. Given the lack of annotations and output schema, more completeness would be helpful for an AI agent.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
With 100% schema description coverage, the input schema already fully documents all 4 parameters. The description adds minimal value beyond what's in the schema - it mentions the optional is_oa filter and pagination (50 results per page), but these are already clearly documented in the parameter descriptions. The baseline score of 3 is appropriate when the schema does the heavy lifting.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the action ('Search Unpaywall for article titles matching a query') and specifies the resource ('article titles'), making the purpose immediately understandable. However, it doesn't explicitly differentiate this tool from its siblings (unpaywall_fetch_pdf_text, unpaywall_get_by_doi, unpaywall_get_fulltext_links), which appear to have different functions but could potentially overlap in some search contexts.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies usage context through 'Search Unpaywall for article titles matching a query' and mentions optional filters/pagination, but doesn't explicitly state when to use this tool versus alternatives like unpaywall_get_by_doi (for DOI-based lookups) or unpaywall_get_fulltext_links (for link retrieval). No explicit when-not-to-use guidance or prerequisite information is provided.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
unpaywall_fetch_pdf_textunpaywall_get_by_doiunpaywall_get_fulltext_linksunpaywall_search_titlesScored across 4 tools
Each tool has a clearly distinct purpose: fetching PDF text, retrieving metadata, getting fulltext links, and searching titles. There is no overlap in functionality, making it easy for an agent to select the right tool for a specific task.
All tools follow a consistent 'unpaywall_verb_noun' pattern, such as unpaywall_fetch_pdf_text and unpaywall_search_titles. This uniformity enhances readability and predictability across the toolset.
With 4 tools, the server is well-scoped for handling Unpaywall-related operations, covering key actions like metadata retrieval, text extraction, link fetching, and title search without being overly sparse or bloated.
The toolset covers essential CRUD-like operations for accessing open-access content, including fetching, searching, and retrieving links. A minor gap might be the lack of update or delete tools, but this is reasonable given the server's read-only nature for external data.
Unpaywall MCP — open-access lookup for scholarly papers (free, no key)
Academic literature search, retrieval, and private library management on top of OpenAlex.
Scrape arXiv, OpenAlex and Crossref papers by author, topic, journal or DOI. Pay per row.
Search arXiv/Semantic Scholar/OpenAlex + medical evidence (PubMed/Europe PMC) + LaTeX/PDF tools.
