Scored across 56 tools
The descriptions are unusually thorough and often explicitly contrast sibling tools (e.g., string_interactions vs experimental_interactions, alphafold_structure vs experimental_structures). However, the set contains several genuinely overlapping clusters: the tair_locus_info/bar_gene_summary alias, multiple synthesis dossiers (gene_report, analyze_locus_synth, biological_context_synth, consensus_homologs), many homology/orthology tools, and the generic batch_locus_call alongside dedicated batch_* variants. An agent could still misselect among these, especially when choosing between a one-shot dossier and its component tools.
Most tools use snake_case with recognizable patterns: batch_* for batch variants, resource_action or locus_attribute for single-source tools, and _synth for prompt-equivalent syntheses. Minor deviations exist, such as verb-first names (get_sequence, resolve_locus_to_uniprot, find_homologs_synth) mixed with noun-first names (locus_go_annotations, interpro_domains, blast_sequence), but the convention remains readable and largely predictable. This is not chaotic, just slightly mixed.
56 tools is far beyond the typical 3-15 range and is inflated by many dedicated batch variants plus synthesis wrappers, even though a generic batch_locus_call already exists. While the server integrates many plant-genomics data sources, the surface feels heavy and contains redundancy that could be consolidated. This is too many tools for an agent to navigate comfortably, despite the breadth of the domain.
The surface covers a wide range of plant-genomics needs: locus resolution, sequences, BLAST, GO and plant ontology, pathways, homology, interactions, coexpression, expression, variation, GWAS, structure, domains, literature, and TF motifs. Minor gaps remain, such as no arbitrary genomic-region sequence retrieval, no multiple sequence alignment or phylogenetic tree tool, and several tools restricted to Arabidopsis or a 12-organism list. Still, the read-only research surface is robust overall.