AIエージェント
arxiv-mcp-server
A local MCP server for serious arXiv work — LaTeX section reads, BibTeX and topic watches.
★ 3.2k スターApache-2.0
ガイドを見る →paper-search-mcpは、openagsが公開しているオープンソースのAIエージェントプロジェクトです（GitHub: openags/paper-search-mcp）。GitHubスターは2,766、フォークは288で、MITライセンスの下で公開されています。GitHubでの最終更新は2026年10月2日です。
MCP server, CLI and skills for searching and downloading academic papers from arXiv, PubMed and 20+ sources
Decide which sources matter for your field (e.g. arXiv for CS, PubMed/Europe PMC for biomedicine) and whether you need search only or PDF download and text extraction too.
Install via uvx/pip/Smithery or clone and run from source.
Add it to your MCP client; optionally set free API keys (Semantic Scholar, CORE, OpenAlex) and the required Unpaywall email in ~/.config/paper-search-mcp/.env.
Ask your client to search_papers for a topic, review the deduplicated results, and download an open-access PDF with the fallback downloader.
Unified search across arXiv, PubMed, bioRxiv, medRxiv, Semantic Scholar, Crossref, OpenAlex, PMC, CORE, Europe PMC, dblp, OpenAIRE, DOAJ, Zenodo, HAL, SSRN, OpenReview, IACR and more.
The high-level search_papers tool runs multi-source concurrent search and deduplicates into a standardized Paper format.
download_with_fallback tries source-native links, then OpenAIRE/CORE/Europe PMC/PMC discovery, then Unpaywall DOI resolution (an optional Sci-Hub step is user-enabled and the user's responsibility).
Works without keys wherever possible; free keys for some sources raise rate limits, and the README documents each source's real capability matrix from live tests.
CLI -s fast / -s fastest modes trade coverage for latency using the quickest sources.
stdio by default, plus SSE and Streamable HTTP bound to localhost, with optional OAuth protected-resource mode for remote deployments.
Clone Paper Search MCP if running from source; most users install it as an MCP server in step 2.
git clone https://github.com/openags/paper-search-mcp.git
cd paper-search-mcpInstall it with the method the README lists for your client (uvx, pip, Smithery, npx or Docker).
# see README for the exact command — install methods include uvx / pip / SmitheryConfigure your client to launch the server (stdio by default), restart the client, and ask it to search for papers.
# see README for the exact command — client config examples are in the Installation sectionopenags maintains it on GitHub.
Yes, MIT-licensed, and deliberately free-first: sources are public/open by default. Some optional free API keys improve rate limits.
No paid key is required. Unpaywall needs just an email set in the env; a few sources (like IEEE) need their own key only if you enable them.
Scholar is one source with bot-detection limits (documented in the README). This server standardizes and deduplicates results across 20+ sources and can download open-access full text.
Install via uvx or Smithery, add it to your MCP client config, restart the client, and ask for a search_papers query.
A local MCP server for serious arXiv work — LaTeX section reads, BibTeX and topic watches.
Exa's MCP server for web search and crawling — give your AI agent a live, high-quality web index.
Brave's official MCP server: web, news, image, video, local, and LLM-context search for your agents.
Bright Data's MCP server: web search, scraping, structured data and browser automation for agents (75 tools)
An MCP server and CLI that lets AI agents build, test and run iOS and macOS projects.
Mobile app automation and verification built for AI coding agents — CLI, MCP server, and Node API
LibreAIカタログに掲載されたAIエージェントプロジェクトの中で最もスターの多いもの — paper-search-mcpが合わなかった場合に検討したい、現実的な候補リストです。
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
カタログの中でpaper-search-mcpとスター数が近いAIエージェントプロジェクト — 前後の棚もそのまま見て回れます。
Recursive self-improving physical agent OS — verified, evidence-based execution for robots and simulators
Terminal security that stops malicious prompts, URLs, and payloads before they run.
Turn papers, text, or topics into editable research figures and slides.
Rebuild Claude Code’s core in ~5,000 lines: an 11-chapter coding-agent masterclass.
Model-agnostic marketplace of agent skills with a harness-free canonical layer.
Chinese Java course: build AI agents with Spring AI, RAG, MCP, and tool calling.