alphaXiv MCP — Semantic ArXiv Search Directly in Claude Code

alphaXiv MCP connects Claude Code directly to arXiv’s millions of papers via the Model Context Protocol. The killer feature: embedding similarity search — it finds papers that are conceptually related to your query, not just keyword matches. Papers using different terminology but addressing the same problem. Papers you’d never find by searching for exact terms. This is the missing piece in AI-assisted literature reviews.

Install — One Command

claude mcp add --transport http alphaxiv https://api.alphaxiv.org/mcp/v1

That’s it. Claude Code now has access to arXiv. No API key needed for basic use.

Known issue: You may need to reactivate the MCP each time you restart Claude Code.

Six Tools Available

Why Embedding Search Changes Research

KEYWORD SEARCH:   "type system" "gradual typing" "security"
                  → finds papers using EXACTLY those terms

EMBEDDING SEARCH: "type systems that improve software security"
                  → finds papers about:
                    - information flow control (different terminology, same concept)
                    - dependent types for memory safety
                    - refinement types for cryptographic protocols
                    - capability-based security (related approach)
                  → papers you'd NEVER find by keyword

The embedding search takes 2-3 sentence descriptions covering concepts, methods, applications, and related terminology. Returns up to 25 papers ranked by relevance + popularity.

Research Workflows

Literature Review (Parallel Search)

Run all three search tools simultaneously for maximum coverage:

1. embedding_similarity_search → conceptual matches
2. full_text_papers_search → exact term matches
3. agentic_paper_retrieval → autonomous deep search
     ↓
   Combine results, deduplicate
     ↓
   get_paper_content → read top papers
     ↓
   answer_pdf_queries → extract specific claims/methods

Reading Dependency Graph

Someone in the community already built this: give a paper, have Claude use alphaXiv to find its key references, then find their references, building a dependency graph of what to read in what order. Perfect for learning a new subfield.

Paper → Code Analysis

get_paper_content → extract GitHub URL from paper
     ↓
read_files_from_github_repository → explore repo structure
     ↓
Claude analyzes implementation details

Alternatives & Comparison

alphaXiv MCP is the easiest to set up (one command, no self-hosting) and the only one with embedding search built in.

How LearnAI Team Could Use This

• Build faster literature reviews — combine semantic, keyword, and agentic search when preparing AI education or coding-agent research briefs.
• Track AI education research — monitor papers on AI-assisted learning, assessment, tutoring, and classroom coding agents.
• Support student research projects — help learners find related work even when they do not know the exact academic terminology.

Real-World Use Cases

For LearnAI Research Areas

For Students

• Literature review assignments — students search for papers semantically, not just by keyword
• Research methodology — teaches the difference between keyword and embedding search
• Reading dependency graphs — Claude builds a “what to read first” map for any topic
• Paper comprehension — answer_pdf_queries lets students ask questions about papers they’re reading

Further Reading

• alphaXiv MCP Documentation
• alphaXiv Explorer
• AlphaSignal: Claude Skill for alphaXiv
• Academix MCP (multi-source alternative)
• Semantic Scholar MCP
• Zotero MCP