Search, download, translate, and summarize academic papers. Use when user asks to find papers, read a paper, translate a PDF, or generate a research briefing with references.
Full workflow: search → download → (translate) → summarize → report.
| Tool | Purpose |
|---|---|
paper_search | Search by keyword/author/topic (Semantic Scholar) |
paper_get | Fetch metadata + abstract for a known paper ID or DOI |
paper_download | Download PDF to workspace/artifacts/papers/ |
translate_pdf | Translate PDF (en→zh or other); returns mono + dual PDF paths |
web_search | Fallback for news, preprints not yet indexed |
When the user asks for a research briefing (e.g. "最新量子计算进展"):
paper_search — 3–5 searches with varied keywords; collect 10–20 candidatesReferences format:
[1] Author et al. (Year). Title. Journal. https://doi.org/...
[2] ...
Never fabricate DOIs or URLs — only use what paper_search/paper_get returns.
# Download (saves to workspace/artifacts/papers/<title>.pdf)
paper_download(paper_id="<semantic_scholar_id_or_doi>")
# Translate to Chinese (requires pdftranslate-mcp)
translate_pdf(file="<absolute_path>", lang_in="en", lang_out="zh")
# Returns: {"mono_pdf": "...", "dual_pdf": "..."}
Use dual_pdf when the user wants to read both languages side-by-side.
Use mono_pdf when the user only wants the translation.
When creating a scheduled briefing (e.g. "每天早上9点发量子计算简报"):
message as a full self-contained prompt including topic, language, format, and reference requirement"Search for the latest quantum computing papers and news from the past 7 days. Write a Chinese briefing with key findings and a References section. Send via feishu."tz="Asia/Shanghai" and cron_exprIf paper_search returns no results, try web_search with site filters:
web_search("quantum computing 2026 site:arxiv.org OR site:nature.com")