Google Scholar MCP Server

<!-- mcp-name: com.hasdata/google-scholar -->

A hosted Model Context Protocol (MCP) server that gives Claude, Cursor, Windsurf and any other MCP client two read-only Google Scholar tools. Search the literature with Scholar's own operators, year ranges and cited-by lookups, then pull a paper's citation in five styles with its BibTeX and EndNote export links, both as structured JSON, with nothing to host.

It reads public Google Scholar pages that a signed-out visitor can see.

1,000 free credits every month, no card required, which is 100 Scholar calls at the 10-credit rate.

https://mcp.hasdata.com/mcp?apis=google_scholar

Contents

• What you need
• Quick start
• Example prompts
• Tools
• Errors and failure paths
• Pricing, free tier and limits
• Tool selection
• How it compares
• FAQ
• HasData links
• Development
• Contributing
• License

What you need

An MCP client and a HasData API key from the dashboard, free to create with no card, and the free tier covers about 100 calls a month at the 10-credit rate. This is a remote server, so the simplest path is a URL and an x-api-key header, with no container to run. A client that only speaks stdio reaches it through a thin launcher, published as @hasdata/google-scholar-mcp on npm and hasdata-google-scholar-mcp on PyPI, shown below.

Quick start

The server URL is the same for every client. We run it hands-on in Claude Code and Claude Desktop. The other blocks follow each client's own documented format for a remote server.

| Field | Value | | :--- | :--- | | URL | https://mcp.hasdata.com/mcp?apis=google_scholar | | Transport | HTTP, streamable | | Auth header | x-api-key: HASDATA_API_KEY |

Clients with OAuth support can add the same URL as a connector and sign in without putting a key in a config file.

<details> <summary><b>Claude Code</b></summary>

claude mcp add --transport http google-scholar "https://mcp.hasdata.com/mcp?apis=google_scholar" \
  --header "x-api-key: HASDATA_API_KEY"

</details> <details> <summary><b>Claude Desktop</b></summary>

Settings, then Connectors, then Add custom connector, then paste https://mcp.hasdata.com/mcp?apis=google_scholar and sign in.

For the config-file route, Claude Desktop loads only local (stdio) servers, so it reaches a remote server through a stdio launcher. The @hasdata/google-scholar-mcp package is that launcher, and it reads the key from the environment. Add this to claude_desktop_config.json:

{
  "mcpServers": {
    "google-scholar": {
      "command": "npx",
      "args": ["-y", "@hasdata/google-scholar-mcp"],
      "env": { "HASDATA_API_KEY": "YOUR_KEY" }
    }
  }
}

For Python instead of Node, swap the launcher for the PyPI package, which uvx runs without a manual install:

{
  "mcpServers": {
    "google-scholar": {
      "command": "uvx",
      "args": ["hasdata-google-scholar-mcp"],
      "env": { "HASDATA_API_KEY": "YOUR_KEY" }
    }
  }
}

</details> <details> <summary><b>Cursor</b></summary>

~/.cursor/mcp.json for every project, or .cursor/mcp.json for one:

{
  "mcpServers": {
    "google-scholar": {
      "url": "https://mcp.hasdata.com/mcp?apis=google_scholar",
      "headers": { "x-api-key": "HASDATA_API_KEY" }
    }
  }
}

</details> <details> <summary><b>Windsurf</b></summary>

~/.codeium/windsurf/mcp_config.json. Windsurf calls the field serverUrl, not url:

{
  "mcpServers": {
    "google-scholar": {
      "serverUrl": "https://mcp.hasdata.com/mcp?apis=google_scholar",
      "headers": { "x-api-key": "HASDATA_API_KEY" }
    }
  }
}

</details> <details> <summary><b>VS Code</b></summary>

.vscode/mcp.json in the workspace:

{
  "servers": {
    "google-scholar": {
      "type": "http",
      "url": "https://mcp.hasdata.com/mcp?apis=google_scholar",
      "headers": { "x-api-key": "HASDATA_API_KEY" }
    }
  }
}

</details>

Example prompts

Each of these lands on one tool, or on two in sequence when the second needs an id the first returns.

• Find papers on transformer architectures published since 2023 and sort them by citation count.
• Who has cited this paper, and how has that grown year on year?
• Give me the BibTeX for this paper.
• Find review articles only on this topic, excluding citations without full records.
• Show me every indexed version of this paper and which of them have a PDF.
• Find recent work by this author on this topic.

A prompt naming a paper takes two calls, one search to reach its resultId and one citation lookup. A prompt about who cites a paper also takes two, because the second call reuses citedBy.citesId from the first.

Tools

| Tool | What it returns | | --- | --- | | hasdata_google_scholar_cite_getScholarCitationFormats | Formatted citation snippets (MLA, APA, Chicago, Harvard, Vancouver) and reference-manager export links (BibTeX, EndNote, RefMan, RefWorks). 10 credits a call | | hasdata_google_scholar_scholar_getScholarSearchResults | Each organic result with title, link, snippet, publication info (authors with profile links), cited-by count and link, related-articles link, and all-versions count and…. 10 credits a call |

Two tools, 10 credits per successful call.

Get Scholar search results

hasdata_google_scholar_scholar_getScholarSearchResults

A page of Scholar results.

| Parameter | Type | Required | Notes | | :--- | :--- | :--- | :--- | | q | string | yes | The query. Scholar operators such as author: and source: work here | | asYlo / asYhi | number | | Published from and up to these years | | start | number | | Result offset, where 0 is the first result | | num | number | | Results per page | | scisbd | number | | 1 sorts abstracts by date, 2 sorts everything by date. Omit for relevance | | cites | string | | Find articles citing this one, using a citedBy.citesId | | cluster | string | | Find every indexed version of an article, using a versions.clusterId | | asSdt | string | | Search type, 0,5 for articles, 4 for case law, 0 or 7 for patents | | asRr | number | | 1 returns review articles only | | asVis | number | | 1 excludes citations, 0 includes them | | hl | string | | Interface language, one of 159 | | lr | array | | Restrict to these content languages | | safe | string | | active or off | | filter | number | | 1 keeps Google's similar and omitted result filters, 0 drops them |

Returns searchInformation with totalResults, queryDisplayed and the time Scholar reported, an organicResults array, and pagination.

Each result carries position, resultId, title, link, snippet, a publicationInfo object, a resources array, citedBy, versions, relatedPagesLink and citeHasdataLink.

publicationInfo.authors is the part worth knowing about. Alongside the raw summary line it lists the authors Scholar has profiles for, each with a name, a profile link and an authorId, which is how you follow one author rather than parsing a byline.

citedBy and versions are the two ids that make this tool compose with itself. citedBy.citesId goes back into cites to walk a citation graph, and versions.clusterId goes into cluster to see every indexed copy of the same paper.

{
  "position": 1,
  "resultId": "A7L9JolPKkoJ",
  "title": "A historical survey of advances in transformer architectures",
  "link": "https://www.mdpi.com/2076-3417/14/10/4316",
  "snippet": "… of the Vision Transformer (ViT) opening a new realm of architectures which build … transformer architecture, it becomes pertinent to examine in detail the architecture of the transformer …",
  "publicationInfo": {
    "summary": "AR Sajun, I Zualkernan, D Sankalpa - Applied Sciences, 2024 - mdpi.com",
    "authors": [
      { "name": "AR Sajun", "authorId": "k6zWX4EAAAAJ", "link": "https://scholar.google.com/citations?user=k6zWX4EAAAAJ&hl=en" }
    ]
  },
  "resources": [{ "fileFormat": "Html", "title": "mdpi.com", "link": "https://www.mdpi.com/2076-3417/14/10/4316" }],
  "citedBy": { "total": 97, "citesId": "5344171358311789059" },
  "versions": { "total": 7, "clusterId": "5344171358311789059" }
}

Get Scholar citation formats

hasdata_google_scholar_cite_getScholarCitationFormats

The citation block for one paper.

| Parameter | Type | Required | Notes | | :--- | :--- | :--- | :--- | | q | string | yes | A resultId from a search result, not a search query | | hl | string | | Interface language |

Returns citations, the formatted string in MLA, APA, Chicago, Harvard and Vancouver, and links, the export URLs for BibTeX, EndNote, RefMan and RefWorks.

{
  "citations": [
    {
      "title": "APA",
      "snippet": "Sajun, A. R., Zualkernan, I., & Sankalpa, D. (2024). A historical survey of advances in transformer architectures. Applied Sciences, 14(10), 4316."
    }
  ],
  "links": [
    { "name": "BibTeX", "link": "https://scholar.googleusercontent.com/scholar.bib?q=info:A7L9JolPKkoJ:scholar.google.com/&output=citation..." }
  ]
}

Errors and failure paths

Plan for these rather than assuming a happy path.

On the citation tool, q is a paper id rather than a query. It takes the resultId from a search result, such as A7L9JolPKkoJ. Passing a title or a DOI there returns nothing useful, and the shared parameter name is the reason people get this wrong.

citesId and clusterId can hold the same value, and they are not interchangeable. They were identical on the paper above. One goes into cites to find papers citing this one, the other into cluster to find copies of this one. Sending the right number to the wrong parameter returns a plausible page of the wrong thing.

type arrives on some results and not others. It was present on one result in five, describing the format of the primary resource. Read resources[].fileFormat when you need to know whether a PDF exists.

A BibTeX link is a Scholar URL with a signature in it, not the BibTeX itself. The links array gives you addresses to fetch, and those carry expiring scisig tokens, so fetch them promptly rather than storing them for later.

totalResults is Scholar's estimate and it is very rough. The query above reported 2,080,000. Treat it as an order of magnitude, never