Dependencies¶

BioMCP is a single Rust binary. All dependencies are compiled in — no runtime installations required.

Article fulltext and conversion stack¶

Article fulltext resolution uses provider APIs and in-process converters rather than external runtime tools. Europe PMC supplies open-access article metadata and fulltext links; NCBI EFetch retrieves PubMed/PMC records; the NCBI ID Converter maps between PMID, PMCID, and DOI when one identifier needs another route; PMC OA supplies open-access package/file discovery; Figshare supports article supplement and hosted-file retrieval; Semantic Scholar PDF links are optional PDF candidates when available. Retrieved content is normalized by Rust transformers: JATS XML through src/transform/article/jats.rs, HTML through src/transform/article/html.rs, and PDF text extraction through src/transform/article/pdf.rs.

Key Dependencies¶

Kuva — Charting¶

Kuva (v0.1.4) is the charting engine behind BioMCP's --chart flag. BioMCP currently exposes 12 chart modes (bar, stacked-bar, pie, waterfall, heatmap, histogram, density, box, violin, ridgeline, scatter, survival) backed by Kuva plots and rendered to terminal, SVG, and PNG.

Kuva is linked as a Rust library, not called as a subprocess. Charts are generated in-process with no additional runtime dependencies.

See the charting guide for examples of every chart type.

RMCP — Model Context Protocol¶

RMCP (v1.1.1) provides the MCP server implementation. BioMCP supports both stdio transport (for local use with Claude Desktop) and streamable HTTP transport (for remote deployment).

Reqwest — HTTP Client¶

Reqwest (v0.12) handles all upstream API calls to biomedical databases. BioMCP adds middleware for:

Retries via reqwest-retry — automatic retry with exponential backoff
Caching via http-cache-reqwest — local HTTP cache to avoid redundant API calls
TLS via rustls — no OpenSSL dependency

Clap — CLI Framework¶

Clap (v4) generates the command-line interface. BioMCP's search, get, study, and chart command hierarchy is defined using Clap's derive macros.

Tokio — Async Runtime¶

Tokio (v1) provides the async runtime. BioMCP uses parallel API fan-out — multiple upstream sources are queried concurrently for faster responses.

Tonic + Prost — gRPC¶

Tonic (v0.12) and Prost (v0.13) support gRPC communication with services that use Protocol Buffers.

MiniJinja — Templating¶

MiniJinja (v2) renders BioMCP's markdown output templates. Each entity type's output format is defined as a Jinja template.

Data Processing¶

Crate	Version	Purpose
`serde` / `serde_json`	1.0	JSON serialization for all API responses
`csv`	1.0	cBioPortal study data parsing (TSV files)
`roxmltree`	0.20	XML parsing for PubMed and ClinicalTrials.gov
`regex`	1.0	Pattern matching for variant notation parsing
`flate2`	1.0	Gzip decompression for downloaded study archives
`tar`	0.4	Tar archive extraction for study downloads
`zip`	0.6	ZIP archive handling
`zstd`	0.13	Zstandard compression

Security¶

Crate	Version	Purpose
`sha2`	0.10	SHA-256 hashing for cache keys
`md5`	0.7	MD5 hashing for content deduplication
`base64`	0.22	Base64 encoding for binary data

Full Dependency List¶

BioMCP's complete dependency tree is defined in Cargo.toml and locked in Cargo.lock.