Summary
Three academic search engines behind one MCP interface. You get arXiv with full-text retrieval in markdown, HTML, or raw LaTeX source; Semantic Scholar's citation graph, author lookup, and recommendations; and OpenAlex's 316M cross-discipline catalog with institution and topic filters. The unified search_all tool fans out to all three, deduplicates by DOI, and re-ranks with reciprocal rank fusion. Bonus: image to LaTeX OCR for turning formula or table screenshots back into source code, backed by DeepSeek-OCR and texify models. Reach for this when you need programmatic paper discovery plus the actual manuscript text, not just metadata. Hosted at latex-tools.online or run locally with FastMCP.
Fastest browser for AI agents to run web automation tasks, always free.
Download Free life-time →
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
belt cli automatically finds the best tools and skills for your agent. image, video, music, tts...
one prompt install →
Connect your Claude agent to live crypto prices and trading routes via 1inch
Get the MCP →
paper-mcp
Remotely-callable MCP server for academic paper search, full-text retrieval & image→LaTeX, served at https://latex-tools.online/mcp.
Three corpora behind one normalized interface:
arxiv(default) — search, metadata, and full-text (HTML / markdown / LaTeX source)semanticscholar(aliass2) — the full S2 API surface: citation graph, authors, recommendations, full-text snippets, bulk datasetsopenalex(aliasoa) — 316M all-field works: citation graph, authors with h-index, institutions, topics, influence metrics
Plus a unified search_all that fuses all three corpora, image→LaTeX OCR, and LaTeX lint + PDF→text tooling.
Tools (41)
Generic / source-agnostic (8)
| Tool | Purpose |
|---|---|
search_all(query, max_results=10, sources='arxiv,semanticscholar,openalex') | Unified search. Fans out to all three corpora concurrently, de-duplicates the same work (by DOI/title) and re-ranks with Reciprocal Rank Fusion. Each hit carries sources (who found it) + an ids map for follow-up calls. Prefer this for broad lookups. |
search_papers(query, source='arxiv', max_results=10, sort_by='relevance') | Single-corpus search. arXiv query accepts plain text or field syntax (ti: au: cat:cs.CL abs: + AND/OR). |
get_paper(paper_id, source='arxiv') | One paper's full record. S2 id accepts S2 id / DOI: / ARXIV: / CorpusId:. |
search_by_author(author, source='arxiv') | Papers by author, newest first. |
list_recent(category, source='arxiv') | Latest in a category (arXiv code or S2 field of study). |
list_categories(source='arxiv') | Common category codes. |
read_paper(paper_id, format='markdown') | FULL text (arXiv). markdown = body with formulas as $LaTeX$; html = raw LaTeXML page; latex = original manuscript .tex source. |
list_paper_sources() | Available corpora. |
read_paper fetch chain: arxiv.org/html/{id} → ar5iv fallback (markdown/html), or arxiv.org/e-print/{id} tarball main .tex (latex). Formulas are recovered from the LaTeXML alttext invariant.
Medical / evidence-graded (1)
| Tool | Purpose |
|---|---|
search_medical(query, study_types='rct,meta-analysis,systematic-review', year_from=0, max_results=10, fetch_fulltext=True) | Clinical literature search. Queries PubMed, filters by research type via Publication-Type tags and re-ranks by the evidence pyramid (meta-analysis / systematic review > RCT > cohort > ...), so real trials surface above high-cited reviews/guidelines that pure-citation ranking floats up. Open-access full text is attached from Europe PMC by PMID. If the type filter yields nothing it auto-relaxes (flagged filter_relaxed). query is English keyword/boolean text — do NL/multilingual query understanding upstream. Backed by NCBI E-utilities + Europe PMC (both free, no key required). |
Image → LaTeX (3)
Turn a formula or table image back into LaTeX (e.g. a figure cropped from a paper) without needing your own vision model. Backed by the co-located recognize service (PaddleOCR-VL / DeepSeek-OCR / texify).
| Tool | Purpose |
|---|---|
recognize_formula(image_url=... or image_base64=..., model='deepseek-ocr') | Formula image → LaTeX. image_url is downloaded server-side (with SSRF guards). Returns {latex, model, elapsed_ms}. |
recognize_table(image_url=... or image_base64=..., model='deepseek-ocr') | Table image → LaTeX tabular. |
list_ocr_models() | Available OCR models (deepseek-ocr, paddleocr-vl, texify). |
LaTeX tooling (3)
Companions to the LaTeX/PDF web tools at latex-tools.online — same backends, exposed over MCP.
| Tool | Purpose |
|---|---|
lint_latex(code) | Check a LaTeX snippet for errors and return an auto-fixed version. Returns {errors, fixed_code, summary_en, summary_zh, elapsed_ms}. |
extract_pdf(pdf_url=... or pdf_base64=..., formula=True, table=True) | PDF → clean Markdown/LaTeX text via MinerU (useful for papers with no open-access full text). pdf_url is downloaded server-side (SSRF-guarded). Content-addressed + cached: a recently-seen or small PDF returns content in one call; a fresh PDF (MinerU is GPU-heavy, minutes) returns status='running' + a task_id. |
extract_pdf_result(task_id) | Fetch an extract_pdf job by task_id. Returns content once status='done'; while 'running', content is null — call again shortly. |
OpenAlex (8)
- Works:
get_openalex_work·get_openalex_citations·get_openalex_references·search_openalex_works(filters: year range, open-access, min-citations, institution) - Authors/Institutions:
search_openalex_authors·search_openalex_institutions - Analytics:
get_openalex_trends·list_openalex_topics
Semantic Scholar (18)
- Graph:
get_paper_citations·get_paper_references·get_paper_authors - Lookup:
match_paper_title·autocomplete_papers - Bulk:
search_papers_bulk(≤1000, sortable, token paging) ·get_papers_batch - Authors:
search_authors·get_author·get_author_papers·get_authors_batch - Full-text:
search_snippets(search inside paper body) - Recommend:
recommend_papers_for_paper·recommend_papers_from_examples - Datasets:
list_dataset_releases·get_dataset_release·get_dataset_download_links·get_dataset_diffs
Layout
paper_mcp/
server.py FastMCP server (tool registrations + instructions)
models.py normalized Paper model
aggregate.py cross-source fusion (dedup + Reciprocal Rank Fusion)
sources/
base.py source registry (get_source / list_sources)
arxiv.py arXiv Atom API + read_paper (HTML/markdown/latex)
semanticscholar.py Semantic Scholar full API surface
openalex.py OpenAlex REST API (works/authors/institutions/topics)
recognize.py image→LaTeX client over the co-located recognize service
latextools.py lint + PDF-extract clients over the latex-tools services
pyproject.toml
Run locally
cd paper-mcp
python -m venv .venv && . .venv/bin/activate
pip install -e .
PAPER_MCP_PORT=9400 python -m paper_mcp.server
# MCP endpoint at http://127.0.0.1:9400/mcp (JSON-RPC; a plain GET returns 406)
Env
| Var | Default | Notes |
|---|---|---|
PAPER_MCP_HOST | 127.0.0.1 | |
PAPER_MCP_PORT | 9400 | |
PAPER_MCP_PATH | /mcp | |
SEMANTIC_SCHOLAR_API_KEY | — | optional; raises S2 rate limit. Set via /etc/paper-mcp.env in prod. |
MCP_MAX_PER_HOUR | 300 | Direct-client JSON-RPC POST budget per IP. |
MCP_WORKER_MAX_PER_HOUR | 300 | Trusted reverse-proxy Worker budget per HMAC-derived connection key. Raw keys are not retained. |
MCP_WORKER_SHARED_MAX_PER_HOUR | 2400 | Shared ceiling across all trusted Worker connections. |
MCP_RATE_COOLDOWN_SEC | 300 | Minimum fast-rejection cooldown after a bucket reaches its limit. |
Deployment (latex-tools.online)
- Runs as
paper-mcp.serviceontencent-us(43.130.32.180), WorkingDirectory/opt/paper-mcp, loopback port 9400. - nginx reverse-proxies
https://latex-tools.online/mcp→127.0.0.1:9400/mcp. - Worker-aware buckets activate only when a trusted reverse proxy overwrites
X-MCP-Workerafter validating the upstream platform. Never pass through a client-supplied value. - uvicorn access logging is disabled because legacy MCP clients may put connection keys and profiles in the endpoint URL. The reverse proxy must also log
$uri, not$request, for the MCP route. - Secrets in
/etc/paper-mcp.env(SEMANTIC_SCHOLAR_API_KEY). - Runtime systemd/nginx/env files are managed by the
tencent-usoperations backup, not by this source repository; never commit/etc/paper-mcp.env.
Update flow
This repo is the source of truth. The server runs an independent copy under /opt/paper-mcp (not auto-synced):
# edit here → push → deploy the complete canonical Python package
rsync -a --delete paper_mcp/ tencent-us:/opt/paper-mcp/paper_mcp/
ssh tencent-us 'systemctl restart paper-mcp'
ssh tencent-us 'curl -s -o /dev/null -w "%{http_code}\n" http://127.0.0.1:9400/mcp' # 406 = healthy (needs JSON-RPC handshake)
Production parity verified on 2026-07-23: main@07f6bbe8622aa063f56ee222a40d19c5d4264048 matches all 12 deployed Python source files byte-for-byte. The older copy embedded in latex-tools-deploy/paper-mcp/ is not a deployment source.
Notes
- arXiv calls are politely rate-limited + retried (
_USER_AGENT, backoff). read_papercovers ~80%+ of papers via official HTML; older scan-only papers may have no full text.- Moved here from the
docsrepo on 2026-06-07; that copy is gone.
License
MIT © MCPServings. See LICENSE.
Featured
Fastest browser for AI agents to run web automation tasks, always free.
Download Free life-time →Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →belt cli automatically finds the best tools and skills for your agent. image, video, music, tts...
one prompt install →Connect your Claude agent to live crypto prices and trading routes via 1inch
Get the MCP →
