paper-search-mcp
Provides tools to search and retrieve preprints in the fields of computer science, mathematics, and physics from the arXiv repository.
Enables DOI lookups and journal metadata retrieval across over 150 million records via CrossRef.
Allows searching and retrieving biomedical literature and research records from the PubMed database.
Enables searching for computer science, social science, and interdisciplinary research papers through the Semantic Scholar platform.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
paper-search-mcp
Academic paper search MCP server for Cloudflare Workers, designed for deployment via mcp-deploy.
Searches across multiple platforms with unified ranking via Reciprocal Rank Fusion (RRF):
Semantic Scholar — CS, social science, interdisciplinary; ML relevance ranking, TLDRs, influential citations, citation graph, paper recommendations
OpenAlex — 250M+ works; FWCI and citation-percentile quality signals, venue metrics, citation graph, semantic search via GTE-Large embeddings
CrossRef — 150M+ records; DOI lookups, journal metadata
arXiv — CS, math, physics preprints
PubMed — Biomedical literature via NCBI E-utilities
bioRxiv / medRxiv — Biology and medical preprints
Core platforms (always on): Semantic Scholar, CrossRef, OpenAlex — provide quality scoring signals (FWCI, h-index, venue quality, citations).
Optional platforms (configurable via ENABLED_PLATFORMS): arXiv, PubMed, bioRxiv, medRxiv. Default: arxiv.
Tools
Tool | Description |
| Unified search across all enabled platforms with RRF fusion. Accepts 1–6 decomposed queries, date filtering, citation thresholds, and blended ranking dials ( |
| Recent-work digest — search scoped to the last N days, optionally filtered to specific journals |
| Deep quality scoring on up to 50 papers (DOIs / S2 IDs / OpenAlex IDs). Fetches rich metadata from S2 + OpenAlex in batched calls, then ranks by age-adaptive quality signals (FWCI, venue impact, citation percentile) |
| ML-powered recommendations from seed papers (positive + optional negative IDs) via Semantic Scholar's embedding recommender; falls back to OpenAlex related works when rate-limited |
| Look up one paper by DOI, arXiv ID, PMID, OpenAlex ID, or S2 ID — auto-detects the ID type and queries the right platform(s) |
| Walk citations (follow-up work) or references (foundational work) of a paper; S2 citation graph with OpenAlex fallback |
Related MCP server: Academic Paper Search MCP Server
Key features
RRF fusion — search results from all platforms are merged with Reciprocal Rank Fusion (score = Σ 1/(k + rank), k=60) and deduplicated by DOI/arXiv ID with metadata merged from the richest source.
Blended ranking — post-fusion scoring blends query relevance, venue/author quality, recency, and citation impact. Four presets (relevance, balanced, discovery, impact) or custom weight dials per call.
Citation graph traversal — get_citation_graph walks references and citations for multi-hop chaining. Backed by Semantic Scholar, with an automatic OpenAlex fallback (cites: filter / referenced_works) that also makes OpenAlex W… IDs resolvable.
Rate-limit survivability — built for good-citizen behavior on free API tiers:
Per-domain throttle that serializes concurrent requests into spaced slots; key-aware spacing for Semantic Scholar (1 req/s with an API key, 1 per 3.5s without)
Exponential backoff with jitter,
Retry-Aftersupport, and a hard per-call time budget60s domain cooldown after a request exhausts its retries on 429, so one hot-limited platform never stalls every search; other platforms keep answering
Batched lookups: one S2
POST /paper/batchand one OpenAlex OR-filter call resolve up to 50 papers inrerank_papersinstead of a request per paperIn-memory TTL response cache (24h for paper metadata, 6h for citation graphs and recommendations, 1h for searches); cache hits are served even while a domain is cooling down
Semantic search — OpenAlex semantic search uses GTE-Large embeddings over 217M works to find conceptually related papers even with different vocabulary. Requires OPENALEX_API_KEY.
Date filtering — search_papers supports date_from / date_to (YYYY-MM-DD), mapped to platform-native filters where possible with post-fusion filtering as a safety net.
Deploy
Install mcp-deploy and deploy to Cloudflare Workers:
npm install -g mcp-deploy
mcp-deploy login
mcp-deploy add upascal/paper-search-mcp
mcp-deploy deploy paper-search-mcpOr use the web UI: mcp-deploy gui
How it works
This repo contains only MCP logic. Auth, deployment, and UI are handled by mcp-deploy (npm install -g mcp-deploy). The repo ships:
src/— MCP server code (Cloudflare Workers + Durable Objects)mcp-deploy.json— deployment contract (secrets, config, worker settings)bench/— retrieval benchmarks (see below)
All platforms return a normalized Paper interface with consistent fields (title, authors, abstract, doi, url, pdf_url, published_date, citations, etc.). Platform-specific data is stored in the extra field.
Configuration
Variable | Required | Description |
| No | Comma-separated list of optional platforms (default: |
| No | Strongly recommended: a dedicated 1 req/s budget instead of the shared anonymous pool. Without it, S2 requests are spaced 3.5s apart and rate-limit cooldowns are common. |
| No* | Enables semantic search ($0.001/query, 100K credits/day free) |
| No | Higher rate limits on E-utilities |
| No | Polite-pool access for CrossRef and OpenAlex (faster, more reliable) |
*OpenAlex keyword search works without a key; semantic search and higher rate limits require one.
Benchmarks
bench/ contains a retrieval evaluation harness over LitSearch (597 queries with ground-truth citations):
npm run bench:download # fetch queries + resolve ground-truth IDs (cached)
npm run bench:run # non-agentic: one federated search per query
npm run bench:eval # compute recall@k / MRR from run output
npm run bench:agentic # agentic: Claude drives the real MCP over stdioRuns are resumable and write JSONL + metrics to bench/litsearch/results/. A SEMANTIC_SCHOLAR_API_KEY in .dev.vars speeds up full sweeps dramatically.
Local development
npm install
npx wrangler dev
# Health check: http://localhost:8787/Testing
npm testUnit tests are offline; integration tests hit live APIs (only S2 rate limiting is tolerated as a skip). Set .dev.vars with optional API keys for reliable runs:
SEMANTIC_SCHOLAR_API_KEY=...
PUBMED_API_KEY=...
OPENALEX_API_KEY=...
CONTACT_EMAIL=you@example.comRelease
Tag a version to trigger the GitHub Actions release workflow:
git tag v0.4.0
git push --tagsThis builds worker.mjs and publishes it alongside mcp-deploy.json as release assets. mcp-deploy fetches these assets to deploy the worker.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Scholarly search: OpenAlex, Crossref, arXiv, OpenCitations and PubMed in one endpoint.
Research papers from OpenAlex, Crossref, arXiv and PubMed in one search, merged by DOI.
1Search 250M scholarly works (OpenAlex, Crossref, arXiv, PubMed) and resolve any DOI
arXiv, PubMed, Crossref and OpenAlex scholarly metadata.
Related MCP Servers
- AlicenseAqualityDmaintenanceEnables users to search, download, and read academic papers from multiple platforms including arXiv, PubMed, bioRxiv, Google Scholar, Semantic Scholar, and CrossRef through a unified interface.347MIT
- FlicenseNot gradedqualityAmaintenanceEnables LLMs such as Claude to search and retrieve academic paper information from Semantic Scholar and Crossref, including metadata, abstracts, full-text content when available, and topic/date-filtered results.-
- AlicenseNot gradedqualityCmaintenanceEnables searching and resolving academic paper metadata across arXiv, Crossref, OpenAlex, Europe PMC, PubMed, and DOAJ, merging duplicate records and returning titles, authors, abstracts, and landing/PDF links. It also supports open-access location lookup via Unpaywall, citation and reference retrieval, author search, open-access full text, and BibTeX/RIS citation export.MIT
- AlicenseAqualityCmaintenanceUnifies nine upstream paper-search servers behind one endpoint, letting an agent search 40+ academic sources, fetch full texts, trace citation graphs, pull peer-review data, and run an end-to-end research pipeline. It auto-detects identifiers like arXiv IDs and DOIs, fans queries out across backends in parallel, and works without any credentials.6MIT