paper-fetch-mcp
paper-fetch-mcp enables AI agents to fetch, resolve, and manage academic papers as structured Markdown and metadata from 18+ publishers and platforms. It operates strictly within the user's existing access permissions and does not bypass paywalls.
resolve_paper— Resolve a DOI, URL, or title query into a normalized paper candidate with confidence scoring, author/title/year match scores, and provider hints, without fetching full content.has_fulltext— Cheaply probe whether a paper likely has accessible full text using metadata and landing-page signals, without downloading the full article.fetch_paper— Fetch a single paper by DOI, URL, or title, returning structured article content, Markdown full text, metadata, references, figures/assets, quality metrics, and token estimates. Supports caching, provider preferences, asset profiles, and optional saving to disk.batch_resolve— Resolve up to 50 DOI/URL/title queries concurrently with shared transport reuse.batch_check— Check availability and metadata for up to 50 papers without returning full bodies.batch_fetch— Fetch up to 50 papers in a single call with bounded concurrency, resumable runs, manifest persistence, and compact output modes.list_cached— List locally cached paper downloads without network access, with options to read the manifest index, refresh/validate it, or rescan from disk.get_cached— Look up DOI-matched cached files for a specific paper, returning entry details, sidecar data, and asset summaries.browser_preflight— Live-check the browser-backed fetch path for browser-dependent providers (Wiley, Science, PNAS, MDPI, etc.), updating stored authentication state as needed.provider_status— Inspect static provider configuration and local dependency readiness (API keys, runtime availability) for any supported publisher, without making live network requests.
Click on "Install 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., "@paper-fetch-mcpfetch full text for DOI 10.1038/s41586-021-03819-2"
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.
Paper Fetch Skill
Fetch papers as agent-ready markdown — DOI/URL/title in, structured full text out. CLI · MCP · Skill.
Paper Fetch Skill —— 已知论文的 AI 阅读层。 你输入 DOI、URL 或标题,它返回结构化元数据 + 干净 Markdown 全文 + 图表资源,直接喂给 Codex / Claude Code / 任意 MCP host。 不绕付费墙,只在你本就有访问权限的地方,把 AI 从「只能读摘要」升级到「读全文」。
如果觉得有帮助,欢迎 star⭐ 支持!
🙁 AI agent 读论文的痛点
你有权限获取全文,但 AI 没有权限,AI 只能搜到摘要。
PDF 无法正确解析文字、图片,agent 理解效果不如 markdown。
文章 html 有很多无关的网页信息,给 agent 造成语义负担和 token 消耗。
文章 html 中的图片 agent 读不到。
Related MCP server: Crossref Academic MCP Server
😍 这个项目做什么
✅这个项目把这些问题收敛到一个工具层:
当你有全文获取权限时,让 AI 也能获取全文,而不仅是摘要。
输入已知论文的 DOI、URL 或标题,抓取 AI 更容易理解的 markdown 版本,为后续知识库构建做好干净的数据基础。
✅项目提供三个主要入口:
paper-fetch:命令行工具,适合手动大规模快速抓取文献。paper-fetch-mcp:stdio MCP server,适合接入 Codex、Claude Code 等支持 MCP 的 host。skills/paper-fetch-skill/:静态 agent skill,告诉 agent 什么时候应该调用论文抓取工具。
核心能力:
支持 DOI、URL 和标题查询。
输出结构化论文元数据、正文 Markdown、引用信息和本地缓存资源。
支持 18 个出版社/平台全文 provider,包括 arXiv、Elsevier、Springer、Wiley、Science、PNAS、IEEE、Copernicus、AMS、MDPI、Royal Society Publishing、Annual Reviews、PLOS、Frontiers、Oxford Academic、ACS、IOP 和 AIP。
在无法取得全文时返回带警告的仅摘要或仅元数据结果。
项目边界:
不替代主题检索、文献推荐或综述生成;开放式搜索可先形成候选,当后续需要阅读、总结、比较、核验可读性或获取全文时,再把 DOI、URL、标题、arXiv ID 或引用条目交给 paper-fetch 抓取和核验候选论文全文。
不绕过付费墙或访问授权;可用性取决于 provider、凭据和本机运行环境。
Wiley、Science、PNAS、Annual Reviews、Royal Society Publishing、ACS、IOP、AIP、MDPI 共用原生 Firefox/Juggler Camoufox browser workflow,见
docs/browser-backends.md。用户可以自行 fork 后添加新出版社,见
onboarding/README.md,但是需要人工审核确定全文获取、markdown 转换质量等能力。
效果展示
agent 安装 skill 后,可以识别 paper-fetch-skill 的适用边界,并在抓取前确认是否保存全文和图表资源。

以下示例来自真实开放抓取产物。
Nature 示例
论文:Towards end-to-end automation of AI research
DOI:
10.1038/s41586-026-10265-5来源:Springer/Nature HTML full text
许可:
CC BY 4.0Markdown 全文:
towards-end-to-end-automation-of-ai-research.md

Science Advances 示例
论文:Deforestation-induced runoff changes dominated by forest-climate feedbacks
DOI:
10.1126/sciadv.adp3964来源:Science Advances / Science provider
Markdown 全文:
deforestation-induced-runoff-changes-dominated-by-forest-climate-feedbacks.md

快速开始
1. 安装
推荐使用 Releases 里的离线安装包:
Windows:下载并运行
paper-fetch-skill-windows-x86_64-setup.exe。Linux:下载匹配 Python ABI 的
paper-fetch-skill-offline-linux-x86_64-cp*.sh。macOS:下载匹配架构和 Python ABI 的
paper-fetch-skill-offline-macos-<arch>-cp*.tar.gz。
每个 v* Release 同时提供 SHA256SUMS、CycloneDX SBOM 和 GitHub build-provenance attestation。
源码安装按能力分层:
python -m pip install . # 轻量 core
python -m pip install ".[browser]" # Camoufox/HTML
python -m pip install ".[pdf]" # PDF
python -m pip install ".[full]" # browser + PDF
uv sync --frozen --extra dev --extra full # 可复现开发环境core 安装包含 MCP Python SDK 2.x(依赖范围 mcp>=2,<3)。paper-fetch-mcp
同时服务 2025 握手协议客户端和 2026-07-28 无状态协议客户端;从源码升级后应重新同步
环境并重启 MCP host。
Windows 安装后新开 PowerShell,验证 CLI:
paper-fetch --help看到 usage: paper-fetch ... 或正常帮助输出即表示 CLI 可用。
Linux 示例:
python3 --version
chmod +x paper-fetch-skill-offline-linux-x86_64-cp312.sh
./paper-fetch-skill-offline-linux-x86_64-cp312.sh --preset=headless --no-user-config
source ~/.local/share/paper-fetch-skill/activate-offline.sh
paper-fetch --helpmacOS 示例:
tar -xzf paper-fetch-skill-offline-macos-arm64-cp312.tar.gz
cd paper-fetch-skill-offline-macos-arm64-cp312
./install-offline.sh --preset=headful --no-user-config
source ~/.local/share/paper-fetch-skill/activate-offline.sh
paper-fetch --help完整安装、升级、卸载和离线包矩阵见 docs/deployment.md。
2. 抓取一篇论文
paper-fetch --query "10.1186/1471-2105-11-421" --output-dir ./papers未显式传 --output 且指定 --output-dir 时,CLI 会把主输出写到该目录,不向 stdout 打印正文。默认文件名使用安全化的论文 stem,优先包含作者、年份和标题;元数据不足时回退 DOI 或标题。需要精确路径时使用 --output ./papers/article.md。
3. 批量抓取
准备 queries.txt:
10.1186/1471-2105-11-421
https://www.nature.com/articles/s41559-026-03039-9运行:
paper-fetch --query-file ./queries.txt \
--output-dir ./papers \
--batch-concurrency 4批量结果会写入 ./papers/batch-results.jsonl,单篇失败会记录后继续后续条目。完整 CLI 输出、artifact、资产和错误码语义见 docs/cli.md。
接入 Agent
Host | 命令 |
Codex |
|
Claude Code |
|
Antigravity CLI |
|
带配置文件注册:
./scripts/install-codex-skill.sh --register-mcp --env-file ~/.config/paper-fetch/.env只安装到当前项目可加 --project。安装后重启对应 host,让它重新扫描 skills 和 MCP 配置。手动 MCP 注册和各 host 路径细节见 docs/deployment.md。
常用配置
默认配置文件位置:
~/.config/paper-fetch/.env创建配置文件:
mkdir -p ~/.config/paper-fetch
cp .env.example ~/.config/paper-fetch/.envElsevier 官方 XML/API 和 PDF fallback 需要从 https://dev.elsevier.com/ 申请 key:
ELSEVIER_API_KEY="..."部分 browser-backed provider 可能需要本机 browser runtime 或手动登录态。默认 Camoufox 在第一次实际抓取时可按需下载 runtime;静态诊断不会下载。需要预检或登录时可运行:
paper-fetch browser-preflight
paper-fetch auth wiley当前 browser-backed auth/preflight provider 包括 wiley、science、pnas、ams、mdpi、royalsocietypublishing、annualreviews、acs、iop、aip。AMS 默认直接启动 Camoufox 尝试站点的静默 JavaScript 验证;已有 provider storage-state 时会自动复用,静默验证失败时可运行 paper-fetch auth ams 完成人工验证并保存状态。完整 provider、运行时和环境变量说明见 docs/providers.md 与 docs/browser-runtime.md。
文档
docs/deployment.md:安装、配置、MCP 注册和更新。docs/browser-backends.md:后端选择、Camoufox runtime、headed 认证、离线准备和 live 验收。docs/cli.md:CLI 输出、artifact、批量抓取和错误码。docs/providers.md:provider 能力、环境变量和运行时配置。docs/README.md:完整文档导航。docs/architecture/overview.md:架构边界和维护者视角。onboarding/README.md:自助添加新 provider。
免责声明
获取的文献仅供个人学术研究和学习使用,不得用于商业用途。
请遵守所在国家/地区著作权法律法规及所在机构的知识产权政策。
本项目不绕过付费墙或访问授权;可用性取决于 provider、凭据和本机运行环境。
本项目不存储、分发或传播任何文献内容,仅协助用户定位、抓取或转换用户有权访问的论文内容。
fixture 中的文献样本仅作为测试使用,严禁对 fixture 进行任何形式的二次分发。
使用者应对自身的文献获取和使用行为承担全部责任。
社区
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- AlicenseBqualityDmaintenanceAn MCP server for searching and downloading academic papers from multiple sources including arXiv, PubMed, bioRxiv, and Sci-Hub, designed for seamless integration with large language models like Claude Desktop.Last updated572,291MIT
- AlicenseAqualityDmaintenanceMCP server enabling AI agents to search and retrieve scientific papers, citations, and author profiles from Crossref, OpenAlex, and Semantic Scholar with no API keys required.Last updated53MIT
- Alicense-qualityBmaintenanceAn MCP server that enables coding agents to search academic papers, ingest full-text PDFs, extract structured details, and manage citations in literature research workflows.Last updated23MIT
- AlicenseAqualityAmaintenanceMCP server for downloading academic papers from DOI or title, resolving references, and generating citations. Supports batch downloads, multiple mirrors, and optional Unpaywall integration.Last updated11MIT
Related MCP Connectors
Academic research MCP server for paper search, citation checks, graphs, and deep research.
OCR, transcription, file extraction, and image generation for AI agents via MCP.
MCP server for AgentDocs (agentdocs.eu): read, search, write, comment on & share Markdown docs.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/Dictation354/paper-fetch-skill'
If you have feedback or need assistance with the MCP directory API, please join our Discord server