Crossref MCP Server
# Crossref MCP Server
A Model Context Protocol (MCP) server for interacting with the Crossref API.
## Features
- Search works by title
- Search works by author
- Get work details by DOI
## Installation
```
{
"mcpServers": {
"crossref": {
"command": "npx",
"args": [
"-y",
"@botanicastudios/crossref-mcp"
]
}
}
}
```
## Usage
The server provides three main tools:
### 1. Search by Title
Search for works in Crossref by title:
```javascript
// Example: Search for works containing "quantum computing" in the title
{
"title": "quantum computing",
"rows": 5 // Optional, defaults to 5
}
```
### 2. Search by Author
Search for works in Crossref by author:
```javascript
// Example: Search for works by "Einstein"
{
"author": "Einstein",
"rows": 5 // Optional, defaults to 5
}
```
### 3. Get Work by DOI
Retrieve a specific work using its DOI:
```javascript
// Example: Get work with DOI "10.1088/1742-6596/1398/1/012023"
{
"doi": "10.1088/1742-6596/1398/1/012023"
}
```
## Response Format
All responses are returned as structured JSON objects with the following format:
### For successful searches:
```json
{
"status": "success",
"query": {
/* the original query parameters */
},
"count": 5,
"results": [
{
"title": "Work title",
"authors": [
{
"given": "First name",
"family": "Last name",
"name": "First name Last name"
}
],
"published": {
"dateParts": [2023, 1, 15],
"dateString": "2023-1-15"
},
"type": "journal-article",
"doi": "10.xxxx/xxxxx",
"url": "https://doi.org/10.xxxx/xxxxx",
"container": "Journal Name",
"publisher": "Publisher Name",
"issue": "1",
"volume": "42",
"abstract": "This is the abstract of the work, if available."
}
// additional results...
]
}
```
### For single DOI lookup:
```json
{
"status": "success",
"query": { "doi": "10.xxxx/xxxxx" },
"result": {
// work details as shown above
}
}
```
### For errors or no results:
```json
{
"status": "error" | "no_results" | "not_found",
"message": "Error message" | null,
"query": { /* the original query parameters */ }
}
```
## Testing
The server comes with a comprehensive test suite using Vitest. Tests cover all available tools and include various scenarios including successful responses, empty results, and error handling.
### Running Tests
```
npm test
```
### Test Structure
The tests use Vitest's mocking capabilities to simulate Crossref API responses without making actual network requests. The test structure includes:
1. **Mock Data**: Sample responses for title searches, author searches, and DOI lookups
2. **Mock Handlers**: Testing versions of the handler functions in `mcp-server-test-handlers.js`
3. **Test Cases**: Tests for all tools covering:
- Successful API responses
- Empty result sets
- Error handling and network failures
### Extending Tests
To add more test cases:
1. Add new mock data to the test file if needed
2. Create additional test cases in the relevant describe block
3. Use the `mockFetchResponse()` helper to simulate API responses
Example:
```javascript
it("should handle a new edge case", async () => {
// Mock the response
mockFetchResponse({
// Your sample response data
});
// Call the handler
const result = await handlers.searchByTitle({ title: "example" });
// Assert the expected results
expect(result).toMatchObject({
// Expected response structure
});
});
```
TDQS
Scored across 3 tools
Each tool has a clearly distinct purpose: retrieving a paper by DOI, searching by author, and searching by title. There is no overlap in functionality, making it easy for an agent to select the appropriate tool based on the query.
The naming follows a consistent verb_noun pattern (getWorkByDOI, searchByAuthor, searchByTitle), with all tools using camelCase. However, the slight inconsistency in 'getWorkByDOI' (using 'Work' instead of 'Paper' or a more specific term) is a minor deviation from perfect alignment with the domain.
With only 3 tools, the server feels thin for a Crossref domain that typically involves broader operations like searching by keyword, year, or journal. While the tools cover basic retrieval and search, the count is borderline low for comprehensive scientific paper access.
The tool surface has significant gaps for a Crossref server. Missing operations include searching by keyword, year, or journal, as well as CRUD-like actions such as listing or filtering results. This incompleteness will likely cause agent failures when handling common academic queries.