[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-72413":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":25,"hasPages":23,"topics":26,"createdAt":10,"pushedAt":10,"updatedAt":37,"readmeContent":38,"aiSummary":39,"trendingCount":16,"starSnapshotCount":16,"syncStatus":40,"lastSyncTime":41,"discoverSource":42},72413,"arxiv-mcp-server","blazickjp\u002Farxiv-mcp-server","blazickjp","A Model Context Protocol server for searching and analyzing arXiv papers","",null,"Python",2843,228,8,4,0,14,40,150,42,109.08,"Apache License 2.0",false,"main",true,[27,28,29,30,31,32,33,34,35,36],"ai","arxiv","claude-ai","gpt","llm","mcp-server","model-context-protocol","papers","python","research","2026-06-12 04:01:05","[![PyPI Version](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Farxiv-mcp-server.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Farxiv-mcp-server\u002F)\n[![PyPI Downloads](https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fdm\u002Farxiv-mcp-server.svg)](https:\u002F\u002Fpypi.org\u002Fproject\u002Farxiv-mcp-server\u002F)\n[![GitHub Stars](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fblazickjp\u002Farxiv-mcp-server?style=flat)](https:\u002F\u002Fgithub.com\u002Fblazickjp\u002Farxiv-mcp-server\u002Fstargazers)\n[![GitHub Forks](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fforks\u002Fblazickjp\u002Farxiv-mcp-server?style=flat)](https:\u002F\u002Fgithub.com\u002Fblazickjp\u002Farxiv-mcp-server\u002Fforks)\n[![Tests](https:\u002F\u002Fgithub.com\u002Fblazickjp\u002Farxiv-mcp-server\u002Factions\u002Fworkflows\u002Ftests.yml\u002Fbadge.svg)](https:\u002F\u002Fgithub.com\u002Fblazickjp\u002Farxiv-mcp-server\u002Factions\u002Fworkflows\u002Ftests.yml)\n[![Python Version](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.11+-blue.svg)](https:\u002F\u002Fwww.python.org\u002Fdownloads\u002F)\n[![License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache%202.0-blue.svg)](https:\u002F\u002Fopensource.org\u002Flicenses\u002FApache-2.0)\n[![smithery badge](https:\u002F\u002Fsmithery.ai\u002Fbadge\u002Farxiv-mcp-server)](https:\u002F\u002Fsmithery.ai\u002Fserver\u002Farxiv-mcp-server)\n[![Install in VS Code](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FInstall_in-VS_Code-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https:\u002F\u002Fvscode.dev\u002Fredirect\u002Fmcp\u002Finstall?name=arxiv-mcp-server&config=%7B%22type%22%3A%22stdio%22%2C%22command%22%3A%22uvx%22%2C%22args%22%3A%5B%22arxiv-mcp-server%22%5D%7D)\n[![Install in VS Code Insiders](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FInstall_in-VS_Code_Insiders-24bfa5?style=flat-square&logo=visualstudiocode&logoColor=white)](https:\u002F\u002Finsiders.vscode.dev\u002Fredirect\u002Fmcp\u002Finstall?name=arxiv-mcp-server&config=%7B%22type%22%3A%22stdio%22%2C%22command%22%3A%22uvx%22%2C%22args%22%3A%5B%22arxiv-mcp-server%22%5D%7D&quality=insiders)\n[![Add to Kiro](https:\u002F\u002Fkiro.dev\u002Fimages\u002Fadd-to-kiro.svg)](https:\u002F\u002Fkiro.dev\u002Flaunch\u002Fmcp\u002Fadd?name=arxiv-mcp-server&config=%7B%22command%22%3A%22uvx%22%2C%22args%22%3A%5B%22arxiv-mcp-server%22%5D%7D)\n[![Codex Plugin](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FCodex-Plugin-412991?style=flat-square)](.\u002F.codex-plugin\u002Fplugin.json)\n\n# ArXiv MCP Server\n\n\u003C!-- mcp-name: io.github.blazickjp\u002Farxiv-mcp-server -->\n\n> 🔍 Enable AI assistants to search and access arXiv papers through a simple MCP interface.\n\nThe ArXiv MCP Server provides a bridge between AI assistants and arXiv's research repository through the Model Context Protocol (MCP). It allows AI models to search for papers and access their content in a programmatic way.\n\n\u003Cdiv align=\"center\">\n  \n🤝 **[Contribute](https:\u002F\u002Fgithub.com\u002Fblazickjp\u002Farxiv-mcp-server\u002Fblob\u002Fmain\u002FCONTRIBUTING.md)** • \n📝 **[Report Bug](https:\u002F\u002Fgithub.com\u002Fblazickjp\u002Farxiv-mcp-server\u002Fissues)**\n\n\u003Ca href=\"https:\u002F\u002Fwww.pulsemcp.com\u002Fservers\u002Fblazickjp-arxiv-mcp-server\">\u003Cimg src=\"https:\u002F\u002Fwww.pulsemcp.com\u002Fbadge\u002Ftop-pick\u002Fblazickjp-arxiv-mcp-server\" width=\"400\" alt=\"Pulse MCP Badge\">\u003C\u002Fa>\n\u003C\u002Fdiv>\n\n## ✨ Core Features\n\n- 🔎 **Paper Search**: Query arXiv papers with filters for date ranges and categories\n- 📄 **Paper Access**: Download and read paper content\n- 📋 **Paper Listing**: View all downloaded papers\n- 🗃️ **Local Storage**: Papers are saved locally for faster access\n- 📝 **Prompts**: A set of research prompts for paper analysis\n\n\n\n## 🔒 Security\n\n### Prompt Injection Risk\n\n**Paper content retrieved from arXiv is untrusted external input.**\n\nWhen an AI assistant downloads or reads a paper through this server, the paper's\ntext is passed directly into the model's context. A maliciously crafted paper\ncould embed adversarial instructions designed to hijack the AI's behavior — for\nexample, instructing it to exfiltrate data, invoke other tools with unintended\narguments, or override system-level instructions. This is a known class of\nattack described by OWASP as **LLM01: Prompt Injection** and by the OWASP\nAgentic AI framework as **AG01: Prompt Injection in LLM-Integrated Systems**.\n\n### Recommended Mitigations\n\n1. **Use read-only MCP configurations** — where possible, configure the MCP\n   client so that the arxiv-mcp-server cannot trigger write operations or invoke\n   other tools on your behalf.\n2. **Review paper content before acting on AI summaries** — if an AI summary\n   asks you to run commands or visit external URLs that were not part of your\n   original request, treat that as a red flag.\n3. **Be cautious in multi-tool setups** — agentic pipelines that combine this\n   server with filesystem, shell, or browser tools are higher risk; a prompt\n   injection in a paper could chain tool calls unexpectedly.\n4. **Treat AI-generated summaries as data, not instructions** — always apply\n   human judgment before executing any action the AI recommends after reading a\n   paper.\n\n### References\n\n- [OWASP LLM01: Prompt Injection](https:\u002F\u002Fowasp.org\u002Fwww-project-top-10-for-large-language-model-applications\u002F)\n- [OWASP Agentic AI - AG01: Prompt Injection](https:\u002F\u002Fgenai.owasp.org\u002Fllmrisk\u002Fag01-prompt-injection\u002F)\n\n---\n\n## 🚀 Quick Start\n\n### Installing via Smithery\n\nTo install ArXiv Server for Claude Desktop automatically via [Smithery](https:\u002F\u002Fsmithery.ai\u002Fserver\u002Farxiv-mcp-server):\n\n```bash\nnpx -y @smithery\u002Fcli install arxiv-mcp-server --client claude\n```\n\n### Installing via Claude Desktop (.mcpb)\n\nThe `.mcpb` bundle is the one-click install path for Claude Desktop on macOS. It bundles the server code and Python package dependencies, so users do not need `uv`, `pip`, or manual MCP JSON configuration. Python 3.11+ must still be available on the user's machine.\n\n1. Download the artifact matching your Mac from the [latest release](https:\u002F\u002Fgithub.com\u002Fblazickjp\u002Farxiv-mcp-server\u002Freleases\u002Flatest):\n   - Apple Silicon: `arxiv-mcp-server-darwin-arm64-\u003Cversion>.mcpb`\n   - Intel: `arxiv-mcp-server-darwin-x86_64-\u003Cversion>.mcpb`\n2. In Claude Desktop open **Settings → Extensions** (or drag-and-drop the file onto the Claude Desktop window).\n3. Click **Install** and, when prompted, set your preferred paper storage directory (defaults to `~\u002F.arxiv-mcp-server\u002Fpapers`).\n\nClaude Desktop launches the bundled server over stdio — no configuration file edits needed.\n\n### Installing Manually\n\n> **Important — use `uv tool install`, not `uv pip install`**\n>\n> Running `uv pip install arxiv-mcp-server` installs the package into the\n> current virtual environment but does **not** place the `arxiv-mcp-server`\n> executable on your `PATH`.  You must use `uv tool install` so that uv\n> creates an isolated environment and exposes the executable globally:\n\n```bash\nuv tool install arxiv-mcp-server\n```\n\nAfter this, the `arxiv-mcp-server` command will be available on your `PATH`.\n\n> **PDF fallback (older papers):** Most arXiv papers have an HTML version which\n> the base install handles automatically. For older papers that only have a PDF,\n> the server needs the `[pdf]` extra (pymupdf4llm). Install it with:\n>\n> ```bash\n> uv tool install 'arxiv-mcp-server[pdf]'\n> ```\nYou can verify it with:\n\n```bash\narxiv-mcp-server --help\n```\n\nIf you previously ran `uv pip install arxiv-mcp-server` and the command is\nmissing, uninstall it and re-install with `uv tool install` as shown above.\n\nFor development:\n\n```bash\n# Clone and set up development environment\ngit clone https:\u002F\u002Fgithub.com\u002Fblazickjp\u002Farxiv-mcp-server.git\ncd arxiv-mcp-server\n\n# Create and activate virtual environment\nuv venv\nsource .venv\u002Fbin\u002Factivate\n\n# Install with test dependencies (development only — no global executable)\nuv pip install -e \".[test]\"\n```\n\n### 🤖 Codex Plugin Integration\n\nThis repository now includes a Codex plugin manifest at `.codex-plugin\u002Fplugin.json`\nand a portable MCP config at `.mcp.json` so Codex-oriented tooling can discover\nthe server without inventing its own install recipe.\n\nThe Codex integration uses the same stdio launch path documented elsewhere in\nthis README:\n\n```json\n{\n  \"mcpServers\": {\n    \"arxiv\": {\n      \"command\": \"uvx\",\n      \"args\": [\"arxiv-mcp-server\"]\n    }\n  }\n}\n```\n\nIf your Codex client supports plugin manifests, point it at\n`.\u002F.codex-plugin\u002Fplugin.json`. If it only supports raw MCP configuration, use\n`.\u002F.mcp.json` directly.\n\n### 🔌 MCP Integration\n\nAdd this configuration to your MCP client config file:\n\n```json\n{\n    \"mcpServers\": {\n        \"arxiv-mcp-server\": {\n            \"command\": \"uv\",\n            \"args\": [\n                \"tool\",\n                \"run\",\n                \"arxiv-mcp-server\",\n                \"--storage-path\", \"\u002Fpath\u002Fto\u002Fpaper\u002Fstorage\"\n            ]\n        }\n    }\n}\n```\n\nFor Development:\n\n```json\n{\n    \"mcpServers\": {\n        \"arxiv-mcp-server\": {\n            \"command\": \"uv\",\n            \"args\": [\n                \"--directory\",\n                \"path\u002Fto\u002Fcloned\u002Farxiv-mcp-server\",\n                \"run\",\n                \"arxiv-mcp-server\",\n                \"--storage-path\", \"\u002Fpath\u002Fto\u002Fpaper\u002Fstorage\"\n            ]\n        }\n    }\n}\n```\n\n### HTTP Transport\n\nFor server deployments where stdio is not practical, run the server with Streamable HTTP:\n\n```bash\nTRANSPORT=http HOST=127.0.0.1 PORT=8080 arxiv-mcp-server --storage-path \u002Fpath\u002Fto\u002Fpapers\n```\n\nThen configure an MCP client that supports Streamable HTTP:\n\n```json\n{\n    \"mcpServers\": {\n        \"arxiv-mcp-server\": {\n            \"type\": \"http\",\n            \"url\": \"http:\u002F\u002F127.0.0.1:8080\u002Fmcp\"\n        }\n    }\n}\n```\n\nThe default HTTP bind host is `127.0.0.1`. Streamable HTTP enables MCP DNS rebinding protection by default and allows loopback hosts for the configured port. If exposing the server through a reverse proxy, keep it bound to localhost unless you have added authentication and network controls upstream; set `ALLOWED_HOSTS` and `ALLOWED_ORIGINS` to the external host\u002Forigin values your proxy forwards.\n\n## 🔒 Security Note\n\narXiv papers are user-generated, untrusted content. Paper text returned by this\nserver may contain prompt injection attempts — crafted text designed to manipulate\nan AI assistant's behavior. Treat all paper content as untrusted input.\n\nIn production environments, apply appropriate sandboxing and avoid feeding raw\npaper content into agentic pipelines that have access to sensitive tools or data\nwithout review. See [SECURITY.md](SECURITY.md) for the full security policy.\n\n## 💡 Available Tools\n\n### Core Workflow\n\nThe typical workflow for deep paper research is:\n\n```\nsearch_papers → download_paper → read_paper\n```\n\n`list_papers` shows what you have locally. `semantic_search` searches across your local collection.\n\n---\n\n### 1. Paper Search\nSearch arXiv with optional category, date, and boolean filters. Enforces arXiv's 3-second rate limit automatically. If rate limited, wait 60 seconds before retrying.\n\n```python\nresult = await call_tool(\"search_papers\", {\n    \"query\": \"\\\"KAN\\\" OR \\\"Kolmogorov-Arnold Networks\\\"\",\n    \"max_results\": 10,\n    \"date_from\": \"2024-01-01\",\n    \"categories\": [\"cs.LG\", \"cs.AI\"],\n    \"sort_by\": \"date\"   # or \"relevance\" (default)\n})\n```\n\nSupported categories include `cs.AI`, `cs.LG`, `cs.CL`, `cs.CV`, `cs.NE`, `stat.ML`, `math.OC`, `quant-ph`, `eess.SP`, and more. See tool description for the full list.\n\n### 2. Paper Download\nDownload a paper by its arXiv ID. Tries HTML first, falls back to PDF. Stores the paper locally for `read_paper` and `semantic_search`.\n\n```python\nresult = await call_tool(\"download_paper\", {\n    \"paper_id\": \"2401.12345\"\n})\n```\n\n> For older papers that only have a PDF, install the `[pdf]` extra: `uv tool install 'arxiv-mcp-server[pdf]'`\n\n### 3. List Papers\nList all papers downloaded locally. Returns arXiv IDs only — use `read_paper` to access content.\n\n```python\nresult = await call_tool(\"list_papers\", {})\n```\n\n### 4. Read Paper\nRead the full text of a locally downloaded paper in markdown. **Requires `download_paper` to be called first.**\n\n```python\nresult = await call_tool(\"read_paper\", {\n    \"paper_id\": \"2401.12345\"\n})\n```\n\n\n\n## 📝 Research Prompts\n\nThe server offers specialized prompts to help analyze academic papers:\n\n### Paper Analysis Prompt\nA comprehensive workflow for analyzing academic papers that only requires a paper ID:\n\n```python\nresult = await call_prompt(\"deep-paper-analysis\", {\n    \"paper_id\": \"2401.12345\"\n})\n```\n\nThis prompt includes:\n- Detailed instructions for using available tools (list_papers, download_paper, read_paper, search_papers)\n- A systematic workflow for paper analysis\n- Comprehensive analysis structure covering:\n  - Executive summary\n  - Research context\n  - Methodology analysis\n  - Results evaluation\n  - Practical and theoretical implications\n- Future research directions\n- Broader impacts\n\n### Pro Prompt Pack\n\n- `summarize_paper`: concise structured summary for one paper.\n- `compare_papers`: side-by-side technical comparison across paper IDs.\n- `literature_review`: thematic synthesis across a topic and optional paper set.\n\n## ⚙️ Configuration\n\nConfigure through command-line options and environment variables:\n\n| Setting | Purpose | Default |\n|---------|---------|---------|\n| `--storage-path` | Paper storage location | `~\u002F.arxiv-mcp-server\u002Fpapers` |\n| `MAX_RESULTS` | Maximum search results | `50` |\n| `REQUEST_TIMEOUT` | API timeout in seconds | `60` |\n| `TRANSPORT` | Transport type: `stdio`, `http`, or `streamable-http` | `stdio` |\n| `HOST` | Host to bind to in HTTP mode | `127.0.0.1` |\n| `PORT` | Port to listen on in HTTP mode | `8000` |\n| `ALLOWED_HOSTS` | Comma-separated extra allowed Host header values for Streamable HTTP DNS rebinding protection | empty |\n| `ALLOWED_ORIGINS` | Comma-separated extra allowed Origin header values for Streamable HTTP DNS rebinding protection | empty |\n\n## 🧪 Testing\n\nRun the test suite:\n\n```bash\npython -m pytest\n```\n\n## 🧪 Experimental Features\n\n> **These features are not yet fully tested and may behave unexpectedly. Use with caution.**\n\nThe following tools require additional dependencies and are under active development:\n\n```bash\nuv pip install -e \".[pro]\"\n```\n\n### Semantic Search\nSemantic similarity search over your **locally downloaded** papers only. Returns empty results if no papers have been downloaded yet. Requires `[pro]` dependencies.\n\n```python\nresult = await call_tool(\"semantic_search\", {\n    \"query\": \"test-time adaptation in multimodal transformers\",\n    \"max_results\": 5\n})\n# or find papers similar to a known paper:\nresult = await call_tool(\"semantic_search\", {\n    \"paper_id\": \"2404.19756\",\n    \"max_results\": 5\n})\n```\n\n### Citation Graph\nFetch references and citing papers via Semantic Scholar. Works on any arXiv ID — no local download required.\n\n```python\nresult = await call_tool(\"citation_graph\", {\n    \"paper_id\": \"2401.12345\"\n})\n```\n\n### Research Alerts\nSave topic watches and poll for newly published papers since the last check. Uses the same query syntax as `search_papers`.\n\n```python\n# Register a watch (idempotent — calling again updates the existing watch)\nawait call_tool(\"watch_topic\", {\n    \"topic\": \"\\\"multi-agent reinforcement learning\\\"\",\n    \"categories\": [\"cs.AI\", \"cs.LG\"],\n    \"max_results\": 10\n})\n\n# Check all watches — returns only papers published since last check\nresult = await call_tool(\"check_alerts\", {})\n\n# Check a single watch\nresult = await call_tool(\"check_alerts\", {\"topic\": \"\\\"multi-agent reinforcement learning\\\"\"})\n```\n\n### Advanced Prompts\n`summarize_paper`, `compare_papers`, and `literature_review` for deeper research workflows. Requires `[pro]` dependencies.\n\n---\n\n## 📄 License\n\nReleased under the Apache License 2.0. See the LICENSE file for details.\n\n---\n\n\u003Cdiv align=\"center\">\n\nMade with ❤️ by the Pearl Labs Team\n\n\u003Ca href=\"https:\u002F\u002Fglama.ai\u002Fmcp\u002Fservers\u002F04dtxi5i5n\">\u003Cimg width=\"380\" height=\"200\" src=\"https:\u002F\u002Fglama.ai\u002Fmcp\u002Fservers\u002F04dtxi5i5n\u002Fbadge\" alt=\"ArXiv Server MCP server\" \u002F>\u003C\u002Fa>\n\u003C\u002Fdiv>\n","ArXiv MCP Server 是一个用于搜索和分析 arXiv 论文的 Model Context Protocol 服务器。该项目通过 MCP 协议为 AI 助手提供了一个访问 arXiv 研究库的桥梁，支持论文检索、内容下载及已下载论文列表查看等功能。采用 Python 编写，并且兼容多种开发环境如 VS Code 和 Kiro 等。它特别适合于需要对学术文献进行自动化处理或集成到更大规模知识管理系统中的场景。",2,"2026-06-11 03:41:55","high_star"]