/ Directory / Playground / Semantic Scholar Skill
← All servers ↗ GitHub
● Community Agents365-ai ⚡ Instant

Semantic Scholar Skill

by Agents365-ai · Agents365-ai/semanticscholar-skill

Semantic Scholar inside Claude — keyword/boolean search, citation traversal, recommendations, authors, batch fetch, with BibTeX/Markdown export.

Agents365-ai/semanticscholar-skill wraps the Semantic Scholar API as a Claude skill. You can search (keyword, boolean, title, snippets, authors), resolve papers by DOI/arXiv/PMID/MAG/ACL/SHA/URL, traverse citations and references, get recommendations (single-seed or positive/negative multi-seed), look up authors (h-index, affiliations, publications), and batch-fetch up to 500 papers / 1000 authors in a single call. Built-in polite rate limiting (1.1s gap, exponential backoff on 429/504). Export to BibTeX, Markdown, or JSON.

▶ Watch demo 📦 Install 💡 Use Cases

Why use it

Key features

Live Demo

What it looks like in practice

semanticscholar-skill.replay ▶ ready
0/0

Install

Pick your client

~/Library/Application Support/Claude/claude_desktop_config.json  · Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "semanticscholar-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/Agents365-ai/semanticscholar-skill.git",
        "~/.claude/skills/semanticscholar-skill"
      ],
      "_inferred": false
    }
  }
}

Open Claude Desktop → Settings → Developer → Edit Config. Restart after saving.

~/.cursor/mcp.json · .cursor/mcp.json
{
  "mcpServers": {
    "semanticscholar-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/Agents365-ai/semanticscholar-skill.git",
        "~/.claude/skills/semanticscholar-skill"
      ],
      "_inferred": false
    }
  }
}

Cursor uses the same mcpServers schema as Claude Desktop. Project config wins over global.

VS Code → Cline → MCP Servers → Edit
{
  "mcpServers": {
    "semanticscholar-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/Agents365-ai/semanticscholar-skill.git",
        "~/.claude/skills/semanticscholar-skill"
      ],
      "_inferred": false
    }
  }
}

Click the MCP Servers icon in the Cline sidebar, then "Edit Configuration".

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "semanticscholar-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/Agents365-ai/semanticscholar-skill.git",
        "~/.claude/skills/semanticscholar-skill"
      ],
      "_inferred": false
    }
  }
}

Same shape as Claude Desktop. Restart Windsurf to pick up changes.

~/.continue/config.json
{
  "mcpServers": [
    {
      "name": "semanticscholar-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/Agents365-ai/semanticscholar-skill.git",
        "~/.claude/skills/semanticscholar-skill"
      ]
    }
  ]
}

Continue uses an array of server objects rather than a map.

~/.config/zed/settings.json
{
  "context_servers": {
    "semanticscholar-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/Agents365-ai/semanticscholar-skill.git",
          "~/.claude/skills/semanticscholar-skill"
        ]
      }
    }
  }
}

Add to context_servers. Zed hot-reloads on save.

claude mcp add semanticscholar-skill -- git clone https://github.com/Agents365-ai/semanticscholar-skill.git ~/.claude/skills/semanticscholar-skill

One-liner. Verify with claude mcp list. Remove with claude mcp remove.

Use Cases

Real-world ways to use Semantic Scholar Skill

Scaffold a literature review around a seed paper

👤 Grad students, researchers starting a review ⏱ ~45 min intermediate

When to use: You have one strong anchor paper and want to build the surrounding literature.

Prerequisites
  • Skill installed — git clone into ~/.claude/skills/semanticscholar-skill
Flow
  1. Resolve the seed
    Resolve 10.1038/s41586-020-2649-2 and get its full metadata.✓ Copied
    → Full paper object with abstract + fields of study
  2. Pull references + citations
    Get the seed's references and top 50 citing papers sorted by influence.✓ Copied
    → Two ranked lists with venues + years
  3. Export Markdown with BibTeX keys
    Export the combined set as Markdown with BibTeX entries I can drop into Zotero.✓ Copied
    → Markdown with inline citekeys + bibtex block

Outcome: A starter lit-review skeleton grounded in real citation data.

Pitfalls
  • Assuming the most-cited paper is the most-relevant paper — Skim abstracts; use recommendations with positive/negative seeds to refine
Combine with: paper-fetch-skill · zotero

Profile an author for an interview, talk, or tenure review

👤 Program chairs, journalists, hiring committees ⏱ ~15 min beginner

When to use: You need a fast, accurate snapshot of someone's publication record.

Flow
  1. Find the author
    Search authors 'Yann LeCun' and pick the NYU/Meta one.✓ Copied
    → Author ID + affiliations + h-index
  2. Get publications
    List their top 50 publications by citation count; group by fields of study.✓ Copied
    → Grouped list with counts; venue + year per paper

Outcome: A defensible author profile with real data.

Pitfalls
  • Disambiguation errors — common names match multiple people — Always inspect affiliation and co-author signal before committing

Get recommendations from what you've liked and disliked

👤 Researchers building a reading list ⏱ ~20 min intermediate

When to use: You have 5 papers you liked and 2 you didn't; find more in that space.

Flow
  1. Provide seeds
    Positive seeds: [DOIs]. Negative seeds: [DOIs]. Recommend 20 papers.✓ Copied
    → Ranked recs biased toward positives, away from negatives
  2. Filter
    Only venues with year ≥ 2022 and open access = true.✓ Copied
    → Filtered recs include only OA post-2022 work

Outcome: A tailored list you'd actually want to read.

Batch-fetch metadata for a long reading list

👤 Anyone importing a bibliography ⏱ ~25 min intermediate

When to use: You have 100+ DOIs and need metadata in one shot.

Flow
  1. Submit the batch
    Batch-fetch metadata for these 200 DOIs; export BibTeX.✓ Copied
    → Single response covering all; BibTeX stream follows
  2. Handle misses
    For IDs that didn't resolve, try match_title as fallback.✓ Copied
    → Fallback search closes most gaps

Outcome: Clean, consistent bibliography ready for Zotero.

Pitfalls
  • Exceeding 500 papers per batch call — Chunk at 500; skill handles rate limits between chunks
Combine with: zotero

Combinations

Pair with other MCPs for X10 leverage

semanticscholar-skill + paper-fetch-skill

Use Semantic Scholar to discover, then fetch the PDFs

Search Semantic Scholar; fetch open-access PDFs for the top 10 hits.✓ Copied
semanticscholar-skill + zotero

Import via Zotero for local library management

Export BibTeX; import into Zotero and tag the collection.✓ Copied
semanticscholar-skill + arxiv

Cross-reference with arXiv metadata

For each arXiv ID in my list, enrich with Semantic Scholar citation counts.✓ Copied

Tools

What this MCP exposes

ToolInputsWhen to callCost
search_relevance query Default paper search 1 request
search_bulk query + filters Bulk results for downstream filtering 1 request
search_snippets query Find papers by phrases in the text 1 request
match_title title Resolve a paper by an imprecise title 1 request
get_paper ID (DOI/arXiv/...) When you have an identifier 1 request
get_citations / get_references paper id, limit Traverse the citation graph 1 request
find_similar / recommend seed(s) (pos + neg) Lookalike and refined recommendations 1 request
batch_papers / batch_authors array of IDs Large reading lists 1 request
search_authors name Profile authors 1 request
export_bibtex / export_markdown / export_json papers[] Final export step 0

Cost & Limits

What this costs to run

API quota
Semantic Scholar public API — rate limited; skill enforces 1.1s gap and exponential backoff
Tokens per call
Metadata is small; abstracts and snippets larger
Monetary
Free — Semantic Scholar is public
Tip
Batch whenever possible — single batch call beats hundreds of get_paper calls.

Security

Permissions, secrets, blast radius

Credential storage: No credentials required for public endpoints. API key optional for higher limits — keep it in env, not config.
Data egress: Queries and results go to api.semanticscholar.org.

Troubleshooting

Common errors and fixes

429 Too Many Requests

Skill already handles this; if persistent, slow your batch pace or request an API key.

Verify: Inspect response headers for x-rate-limit-*
Paper not found by DOI

Fall back to match_title with the paper's title; ~10% of papers lack DOI coverage.

Verify: match_title + grep output
Author disambiguation wrong

Filter by affiliation or co-author before committing; Semantic Scholar occasionally merges distinct people.

Verify: Inspect the author's publication list for coherence

Alternatives

Semantic Scholar Skill vs others

AlternativeWhen to use it insteadTradeoff
arxivYou only care about arXiv, not the broader citation graphNarrower coverage; no citation traversal
paper-fetch-skillYou have IDs and want PDFs more than metadataFetcher, not a search engine
asta-skillYou want a different academic search flavorDifferent source

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills