paper-fetch MCP — 使用场景, 安装 & 实时演示

为什么要用

核心特性

DOI → PDF resolution across 4+ legal sources
Unpaywall integration for OA detection
arXiv / bioRxiv / PMC direct fetching
Multi-platform (Claude Code, generic Agent Skill)
Clear licensing respect — no sci-hub, no pirated content

实时演示

实际使用效果

paper-fetch-skill.replay ▶ 就绪

0/0

安装

选择你的客户端

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "paper-fetch-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/Agents365-ai/paper-fetch",
        "~/.claude/skills/paper-fetch"
      ],
      "_inferred": true
    }
  }
}

打开 Claude Desktop → Settings → Developer → Edit Config。保存后重启应用。

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "paper-fetch-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/Agents365-ai/paper-fetch",
        "~/.claude/skills/paper-fetch"
      ],
      "_inferred": true
    }
  }
}

Cursor 使用与 Claude Desktop 相同的 mcpServers 格式。项目级配置优先于全局。

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "paper-fetch-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/Agents365-ai/paper-fetch",
        "~/.claude/skills/paper-fetch"
      ],
      "_inferred": true
    }
  }
}

点击 Cline 侧栏中的 MCP Servers 图标，然后选 "Edit Configuration"。

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "paper-fetch-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/Agents365-ai/paper-fetch",
        "~/.claude/skills/paper-fetch"
      ],
      "_inferred": true
    }
  }
}

格式与 Claude Desktop 相同。重启 Windsurf 生效。

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "paper-fetch-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/Agents365-ai/paper-fetch",
        "~/.claude/skills/paper-fetch"
      ]
    }
  ]
}

Continue 使用服务器对象数组，而非映射。

~/.config/zed/settings.json

{
  "context_servers": {
    "paper-fetch-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/Agents365-ai/paper-fetch",
          "~/.claude/skills/paper-fetch"
        ]
      }
    }
  }
}

加入 context_servers。Zed 保存后热重载。

claude mcp add paper-fetch-skill -- git clone https://github.com/Agents365-ai/paper-fetch ~/.claude/skills/paper-fetch

一行命令搞定。用 claude mcp list 验证，claude mcp remove 卸载。

使用场景

实战用法： paper-fetch

Pull all PDFs for a literature review reading list

👤 Researchers, grad students, science writers ⏱ ~15 min beginner

何时使用： You have 40 DOIs from a literature search and want to skim them without 40 browser clicks.

前置条件

Skill installed — git clone https://github.com/Agents365-ai/paper-fetch ~/.claude/skills/paper-fetch
Unpaywall email for API access — Free; email is the API key: https://unpaywall.org/products/api

步骤

Hand over the list

Use paper-fetch. Here are 40 DOIs. Download the OA version of each to ~/papers/lit-review/.✓ 已复制

→ Progress per DOI; summary of found/unfound
Handle misses

For ones not found, tell me what's available (abstract-only? preprint missing?).✓ 已复制

→ Status per miss
Index the results

Make an index.md with title, year, and file path per downloaded paper.✓ 已复制

→ Browsable index

结果： A local folder of PDFs you can actually search.

注意事项

Some papers aren't legally OA anywhere — Skill reports the gap; request via your library's ILL for those

搭配使用： filesystem

Chase citations one hop at a time

👤 Deep researchers following reference chains ⏱ ~20 min intermediate

何时使用： You found a good paper and want the most-cited works it references.

步骤

Start with the seed

Use paper-fetch. Download this paper [DOI] and list its top-10 most-cited references.✓ 已复制

→ Paper downloaded; referenced DOIs listed
Fetch the next ring

Download those 10 too.✓ 已复制

→ Batch download

结果： A mini research trail in your file system.

搭配使用： arxiv

组合

与其他 MCP 搭配，撬动十倍杠杆

paper-fetch-skill + arxiv

Prefer arXiv for preprints, fall back to Unpaywall for journal OA

Prefer arXiv preprints when present; otherwise use Unpaywall.✓ 已复制

paper-fetch-skill + filesystem

Organized local library

Save to ~/papers/<year>/<first-author>/✓ 已复制

paper-fetch-skill + infinite-gratitude-skill

Multi-agent research with real PDFs

Dojo explorer uses paper-fetch to ground claims in actual papers.✓ 已复制

工具

此 MCP 暴露的能力

工具	输入参数	何时调用	成本
fetch_by_doi	DOI	Per paper	1-3 API lookups
batch_fetch	DOI list	Lit reviews	N * lookups
resolve_citation	free-text citation	Starting from citation text	Crossref + Unpaywall

成本与限制

运行它的成本

API 配额: Unpaywall 100k/day (with email); arXiv and PMC are generous
每次调用 Token 数: Low
费用: Free
提示: Batch DOIs; respect rate limits to avoid being blocked.

安全

权限、密钥、影响范围

凭据存储： Unpaywall email (not a secret) in env var

数据出站： Unpaywall, arXiv, PMC, bioRxiv APIs only

This skill only retrieves legally open-access content. Do not modify it to bypass paywalls; respect publisher licensing.

故障排查

常见错误与修复

PDF not found for a DOI you know is OA

Unpaywall occasionally lags; check oadoi.org directly or wait a day

验证： curl https://api.unpaywall.org/v2/<DOI>[email protected]

Rate limited

Space batches out; use the skill's rate-limit-aware batch mode

Wrong PDF downloaded (supplementary instead of main)

Some sources return supplements; rerun with prefer-main flag

替代方案

paper-fetch 对比其他方案

替代方案	何时用它替代	权衡
arxiv MCP	You only need arXiv, nothing else	Narrower scope
Manual browser + Zotero	Small reading lists	Tedious at scale

paper-fetch