# doiget [Health: Active]

**Category:** 🔬 Research  
**Repository:** https://github.com/QAtlasHub/doiget  
**GitHub Stars:** 2  
**npm Downloads (last month):** 240  
**Views:** 2  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/doiget-2

## Description
OA-first paper fetcher: DOIs/arXiv IDs to local PDFs + metadata via Open Access APIs.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "doiget": {
    "command": "npx",
    "args": ["-y","doiget-cli"]
  }
}
```

## Documentation

## What the doiget MCP server does

The doiget MCP server connects agent workflows to paper-retrieval services for DOI and arXiv identifiers. It resolves those identifiers, checks official open-access sources, and writes available papers as local PDF files alongside metadata. The project is distributed as a single binary that can operate as either a command-line program or a stdio MCP server.

Its retrieval model is open-access first. The documented sources include Crossref, Unpaywall, and arXiv. The project also implements citation-graph expansion and gated text-and-data-mining sources, although access to those paths depends on the relevant configuration and source requirements.

## How it works

An MCP host starts the binary and communicates with it over standard input and output. An agent can then provide a DOI or arXiv ID for retrieval through the server interface. doiget consults official APIs and returns or records the resulting local file and metadata rather than acting as a content-hosting service.

The project supports two access modes: public open-access retrieval by default, and credentials that a user has personally configured for institutional or personal subscriptions when the corresponding capability is enabled at compile time. It does not attempt to defeat authentication, evade paywalls, redistribute papers, or provide a hosted SaaS endpoint.

## Setup and configuration

Install a prebuilt release binary with the shell installer on Linux or macOS, the PowerShell installer on Windows, Homebrew, or a platform package. The project also documents `cargo install doiget-cli`, which builds the `doiget` binary from source, and `npx -y doiget-cli serve` for starting the MCP server without a separate installation step.

For Claude Code, the repository provides a plugin whose MCP configuration launches the npm package. The README also identifies Claude Desktop as an installation target through a `.mcpb` extension. Installer settings include `DOIGET_INSTALL_DIR` for changing the shell installer destination and `DOIGET_VERSION` for selecting a release; neither is required for normal server operation.

## Tools and capabilities

The doiget MCP server is suited to agent tasks that need to:

- Resolve DOI and arXiv identifiers.
- Retrieve legally available paper PDFs.
- Obtain associated bibliographic metadata.
- Expand through citation relationships.
- Check the running installation through the documented health and capability commands.

The command-line interface includes `doiget capabilities` and `doiget_health` reporting for installation and version details. These reports can identify the active version, release channel, installation method, and the associated update command.

## Limitations and notes

Retrieval depends on whether an official open-access copy is available through the supported sources. A DOI alone does not guarantee that a PDF can be downloaded. Users remain responsible for confirming that their access and requested use comply with source terms and applicable rights.

The doiget MCP server is local rather than hosted: it does not operate as a SaaS service or store a central paper collection. It also does not bundle publisher API keys. Subscription-based retrieval is opt-in and compile-time gated, so deployments should verify that the selected binary and configuration support the required source.

The Linux release is fully static and uses musl, which is intended to help it run on older glibc systems and HPC environments. Release binaries are checksum-verified, and the project states that it does not update itself automatically.

_Full upstream README: https://allmcps.com/mcp/doiget-2/readme_