Scored across 50 tools
Several tools occupy overlapping functional territory: tair_locus_info is literally an alias of bar_gene_summary, and locus_variants, arabidopsis_natural_variation, and aragwas_associations all return variant data for a locus. The many batch_* duplicates and synthesis wrappers such as gene_report, analyze_locus_synth, and biological_context_synth further blur boundaries, even though individual descriptions often clarify the intended use case.
Names are consistently snake_case and group into recognizable families like locus_*, get_*, and batch_*, but the set mixes verb-first names (get_sequence, resolve_locus_to_uniprot), resource-first names (locus_literature, string_interactions), and source-first names (bar_gene_summary, phytozome_lookup_locus). It is readable and searchable, but there is no single predictable verb_noun convention across the server.
Fifty tools is excessive for one MCP server, especially because 12 are batc_* duplicates of existing single-locus tools and several synthesis wrappers overlap with one another. Even for broad plant genomics, this surface would be far easier to navigate as multiple focused servers or with batc functionality merged into the base tools.
The server covers an impressively broad read-only plant genomics workflow: sequence retrieval, BLAST, GO and ontology annotations, pathways, structures, variants, interactions, homology, motifs, literature, and synthesis reports. Minor gaps exist—such as no tissue/developmental expression atlas or synteny browser—but agents can work around them using the existing tools and batch/synthesis endpoints.