Gbif Biodiversity MCP Server

What gbif-biodiversity-mcp-server does

GBIF taxonomy and biodiversity data become available to MCP-compatible agents through 13 tools. The server covers four related areas:

• Species identification and taxonomy, including scientific-name matching, classifications, synonyms, descendants, and taxon browsing.
• Occurrence data, including filtered searches, record counts, full Darwin Core records, and grouped facets.
• Dataset metadata, such as citations, contacts, licenses, DOIs, coverage, and publishing organizations.
• GBIF-registered publishers, searchable by organization name or country.

The taxonomy tools can resolve a name to a GBIF taxon key, including bulk matching for up to 50 names. Those keys can then be used with occurrence queries or taxon navigation tools. Species resources are available at gbif://species/{taxonKey}, while dataset resources use gbif://dataset/{datasetKey}.

How it works

Requests are exposed over MCP using either STDIO or Streamable HTTP. The public hosted endpoint is https://gbif-biodiversity.caseyjhand.com/mcp. A client can also run the package locally through its MCP configuration.

Species searches can use a name fragment, rank, extinction status, checklist dataset, or taxonomic scope such as kingdom, family, or genus. Matching returns a taxon key, confidence, match type, and classification. Synonym matches include the accepted taxon information, while higher-rank matches identify the broader taxon that GBIF resolved.

Occurrence searches accept filters including taxon, place, date, basis of record, presence or absence, and IUCN category. Facets aggregate matching records by dimensions such as country, year, dataset, taxon, and status. Dataset and publisher tools help connect observations with their source and organizational context.

Setup and configuration

The repository provides direct installation paths for Claude Desktop, Cursor, and VS Code. Cursor and VS Code configurations use the npm package through npx, while Claude Desktop has a downloadable .mcpb package. The package is published as @cyanheads/gbif-biodiversity-mcp-server.

The server can also be reached through the hosted Streamable HTTP endpoint instead of running a local process. Docker packaging is available through the project’s GitHub Container Registry listing.

Tools and capabilities

• Match one scientific name or bulk-match up to 50 names.
• Fetch a taxon’s classification, authorship, synonymy, vernacular name, parent, and descendant count.
• Search the GBIF backbone by name, rank, kingdom, family, or genus.
• List a taxon’s parent chain or direct children.
• Search more than 3.9 billion occurrence records with GBIF filters.
• Count matching occurrences without retrieving records.
• Retrieve a complete occurrence by key.
• Aggregate occurrence results into facets, optionally resolving taxon and dataset names.
• Search datasets and retrieve full metadata by UUID.
• Search registered publishers by name or country.

Limitations and notes

Bulk matching processes the first 50 names in input order. When more names are supplied, the remaining names are returned for a later request when read-limit clamping is enabled.

A taxon record does not provide an occurrence count; use gbif_count_occurrences for that purpose. Species-search rows can contain a placeholder zero for numOccurrences, so occurrence counts should be obtained through an occurrence count query instead.

Some MCP clients expose tools but not resources. The species and dataset resources are also reachable through gbif_get_species and gbif_get_dataset, allowing tool-only clients to access the same underlying records.