# Scite Research Assistant

> Search and analyze scientific literature using scite's MCP server. Use when users ask about scientific research, want to find papers, verify citations, fact-check scientific claims, check for retractions, or need evidence from peer-reviewed literature.

- Skill: `scitedotai/scite-research-assistant` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add scitedotai/scite-research-assistant`
- Raw SKILL.md: https://api.skillmd.com/api/skills/scitedotai/scite-research-assistant/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: scitedotai (https://skillmd.com/u/scitedotai)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/scitedotai/scite-research-assistant

---


# scite Research Assistant

Guide Claude to effectively search and analyze scientific literature using scite's `search_literature` MCP tool.

## When to Use This Skill

Activate when users:
- Ask about scientific research, studies, or evidence
- Want to find papers on a topic
- Need to verify or fact-check scientific claims
- Ask to check references or DOIs
- Want to know how papers cite each other
- Need citation metrics or retraction checks
- Ask for literature reviews or evidence summaries

## Core Tool: `search_literature`

The scite MCP server exposes one tool with many parameters. Use them strategically.

### Basic Search

For general topic queries, use `term` with relevant keywords:

```
search_literature(term="cognitive behavioral therapy insomnia efficacy")
```

### DOI Lookup

To retrieve a specific paper, use `doi`:

```
search_literature(doi="10.1038/s41586-020-2012-7")
```

### Filtered Search

Combine filters for precise results. Available filters include:

| Parameter | Description | Example |
|-----------|-------------|---------|
| `term` | Search query across title, abstract, fulltext | `"CRISPR gene therapy"` |
| `doi` | Specific paper DOI | `"10.1038/s41586-020-2012-7"` |
| `author` | Author name | `"John Smith"` |
| `journal` | Journal name | `"Nature"` |
| `topic` | Research topic/field | `"machine learning"` |
| `affiliation` | Institution | `"MIT"` |
| `date_from` / `date_to` | Publication year range | `2020` / `2024` |
| `publication_type` | Type of publication | `"journal-article"` |
| `citing_publications_from` | Min citing papers | `50` |
| `supporting_from` | Min supporting citations | `10` |
| `contrasting_from` | Min contrasting citations | `3` |
| `mentioning_from` | Min mentioning citations | `5` |
| `limit` | Number of results (default 10) | `20` |
| `sort` | Sort order | `"relevance"`, `"date"` |

### Reading Full Text

For open access papers, scite returns `fulltextExcerpts` — relevant passages from the paper matching your search term. To read different sections, make multiple calls with different terms:

```
search_literature(doi="10.1038/s41586-020-2012-7", term="methods")
search_literature(doi="10.1038/s41586-020-2012-7", term="results")
search_literature(doi="10.1038/s41586-020-2012-7", term="discussion")
```

## Understanding Results

Each result includes:

- **Title, authors, journal, year** — Standard metadata
- **Abstract** — Paper summary
- **Smart Citations** — Snippets from other papers that cite this one, classified as:
  - **Supporting** — Cites the paper in agreement
  - **Contrasting** — Cites the paper in disagreement or with conflicting findings
  - **Mentioning** — Neutral reference
- **Citation counts** — `supportingCount`, `contrastingCount`, `mentioningCount`, `citingPublicationsCount`
- **Retraction/correction status** — `editorialNotices` flags retractions, corrections, expressions of concern
- **Full-text excerpts** — Passages from open access papers matching the search term
- **Access links** — Resolved URLs with open access, institutional, and purchase options via Article Galaxy

## Recommended Workflows

### Literature Review

1. Search broadly: `search_literature(term="topic keywords", limit=20)`
2. Identify key papers from results
3. Deep-dive into top papers: `search_literature(doi="...", term="methods")`
4. Check citation patterns — look at supporting vs contrasting counts
5. Synthesize findings with proper citations

### Fact-Checking a Claim

1. Find the source paper: `search_literature(doi="...")`
2. Check for retractions in `editorialNotices`
3. Search for related work: `search_literature(term="claim keywords")`
4. Compare Smart Citations — do citing papers support or contrast the claim?
5. Present balanced evidence assessment

### Reference Verification

1. Look up each DOI: `search_literature(doi="...")`
2. Verify the paper exists and metadata matches
3. Check `editorialNotices` for retractions or corrections
4. Compare claimed findings against actual abstract and Smart Citations
5. Flag any discrepancies

### Finding Debated Topics

1. Search with contrasting filter: `search_literature(term="topic", contrasting_from=3)`
2. Review the contrasting citation snippets to understand disagreements
3. Search for highly-cited papers: `search_literature(term="topic", citing_publications_from=50)`
4. Cross-reference to find influential but contested work

## Citation Format

Always cite papers using standard academic format. Preferred: APA style.

Example:
> Smith, J., & Doe, A. (2023). Title of paper. *Journal Name*, 45(2), 123-135. https://doi.org/10.xxxx/xxxxx

When presenting multiple results, include a numbered reference list at the end.

## Guidelines

- Always check `editorialNotices` before citing a paper — warn users about retracted papers
- Present both supporting and contrasting evidence for balanced assessments
- Use `fulltextExcerpts` when available to provide direct quotes from papers
- When a paper has high contrasting citations, highlight the debate
- Do not fabricate citations — only cite papers returned by scite
- If a DOI lookup returns no results, tell the user the paper was not found rather than guessing
- Use multiple targeted searches rather than one broad search for complex questions