# Available MCP Tools
Once connected via MCP, the following **unified tools** are available. These tools automatically select the best source based on your input, or you can specify a source explicitly.
## Core Research Tools
### search_papers
Search for papers across all available research sources or a specific source.
**Parameters:**
- `query` (required): Search query string
- `source` (optional): Specific source to search (e.g., "arxiv", "semantic", "pubmed")
- `max_results` (optional): Maximum number of results (default: 10)
- `year` (optional): Year filter (e.g., "2020", "2018-2022", "2010-", "-2015")
- `category` (optional): Category/subject filter
**Example:**
```json
{
"query": "transformer architecture",
"year": "2020-",
"max_results": 20
}
```
### search_by_author
Search for papers by a specific author across sources that support author search.
**Parameters:**
- `author` (required): Author name
- `source` (optional): Specific source to search
- `max_results` (optional): Maximum results per source (default: 10)
**Supported sources:** arxiv, semantic, openalex, pubmed, biorxiv, pmc, hal, iacr, ssrn
### get_paper
Get detailed metadata for a specific paper. The source is auto-detected from the paper ID format.
**Parameters:**
- `paper_id` (required): Paper identifier (e.g., "2301.12345", "arXiv:2301.12345", "PMC12345678")
- `source` (optional): Override auto-detection and use specific source
**Auto-detection patterns:**
- `arXiv:xxxx` or `xxxx.xxxx` (numeric) → arXiv
- `PMCxxxxxxx` → PMC
- `10.xxxx/xxxxxx` (DOI) → Source with DOI lookup capability
- `CORX:xxxxx` → CORE
- `hal-xxxxxxx` → HAL
- `xxxx/xxxx` (IACR format) → IACR
- `DOI:10.xxxx` → DOI-based lookup
- etc.
### download_paper
Download a paper PDF to your local filesystem.
**Parameters:**
- `paper_id` (required): Paper identifier
- `source` (optional): Override auto-detection
- `output_path` (optional): Save path (default: ./downloads)
- `auto_filename` (optional): Auto-generate filename from title (default: true)
### read_paper
Extract and return the full text content from a paper PDF.
**Parameters:**
- `paper_id` (required): Paper identifier
- `source` (optional): Override auto-detection
**Note:** Requires poppler to be installed. Returns an error if PDF extraction fails.
## Citation & Reference Tools
### get_citations
Get papers that cite a specific paper. Prefers Semantic Scholar for best results.
**Parameters:**
- `paper_id` (required): Paper identifier
- `source` (optional): Specific source (default: "semantic")
- `max_results` (optional): Maximum results (default: 20)
### get_references
Get papers referenced by a specific paper. Prefers Semantic Scholar.
**Parameters:**
- `paper_id` (required): Paper identifier
- `source` (optional): Specific source (default: "semantic")
- `max_results` (optional): Maximum results (default: 20)
## Lookup Tools
### lookup_by_doi
Look up a paper by its DOI across all sources that support DOI lookup.
**Parameters:**
- `doi` (required): Digital Object Identifier (e.g., "10.48550/arXiv.2301.12345")
- `source` (optional): Specific source to query (default: all)
**Supported sources:** semantic, openalex, crossref, hal, doaj, osf, springer, mdpi, acm, base, unpaywall
## Utility Tools
### deduplicate_papers
Remove duplicate papers from a list using DOI matching and title similarity.
**Parameters:**
- `papers` (required): Array of paper objects
- `strategy` (optional): Deduplication strategy - "first" (keep first), "last" (keep last), or "mark" (add `is_duplicate` flag)
**Deduplication criteria:**
- Exact DOI match
- Title similarity > 0.95 (Jaro-Winkler algorithm)
- Author verification
## Smart Source Selection
The unified tools use intelligent source auto-detection:
| `arXiv:1234.5678` or `1234.5678` | arXiv |
| `PMC12345678` | PMC |
| `10.xxxx/xxxxxx` | Source with DOI lookup |
| `CORX:xxxxx` | CORE |
| `hal-xxxxxxx` | HAL |
| `xxxx/xxxx` (IACR format) | IACR |
| `doi:10.xxxx` or `https://doi.org/10.xxxx` | DOI lookup |
| `ZENODO:xxxxx` | Zenodo |
| etc. |
You can always override auto-detection by specifying the `source` parameter explicitly.
## Example Usage with Claude
Once configured with Claude Desktop, you can interact with the research sources naturally using the unified tools:
```
User: Search for papers about "transformer architecture" from 2020 onwards
Claude: [Uses search_papers tool with query="transformer architecture", year="2020-"]
User: Find papers by Geoffrey Hinton on deep learning
Claude: [Uses search_by_author tool with author="Geoffrey Hinton"]
User: Download the paper "Attention Is All You Need" and find what papers cite it
Claude: [Uses download_paper with paper_id="1706.03762" (arXiv),
then get_citations to find citing papers]
User: Read the abstract and introduction from this paper
Claude: [Uses read_paper tool with paper_id="1706.03762" to extract PDF text]
User: Look up this paper by its DOI
Claude: [Uses lookup_by_doi with doi="10.48550/arXiv.1706.03762"]
```
The unified tools automatically detect the appropriate source, so you don't need to remember which source has which paper. You can simply provide the paper ID, DOI, or search query, and the tool handles the rest.
## Related Documentation
- [Sources](sources.md) - Supported research sources
- [Usage](usage.md) - CLI commands and options
- [MCP Clients](mcp-clients.md) - Client configuration
- [Configuration](configuration.md) - Environment variables