Scored across 58 tools
The descriptions work hard to separate near-neighbors (e.g. 'Unlike locus_go_annotations...', 'Unlike string_interactions...'), which rescues much of the overlap. But the surface still has real duplication: tair_locus_info is explicitly 'Alias of bar_gene_summary', five synthesis tools (gene_report, analyze_locus_synth, biological_context_synth, consensus_homologs, find_homologs_synth) partially re-run the same fan-outs, and 15+ dedicated batch_* tools are subsumed by batch_locus_call.
A source-prefix convention (ensembl_plants_, bar_, batch_, locus_) is applied fairly consistently, but verb style is mixed: resource-first nouns (gramene_homologs, kegg_pathways, interpro_domains, tf_binding_motifs) sit alongside verb_noun names (get_sequence, resolve_locus_to_uniprot, blast_sequence, vep_annotate, entry_members). Suffix conventions also split (_synth suffix vs sync tools, batch_ prefix vs no batch variant for some tools), so it is readable but not predictable.
58 tools is far past a well-scoped surface for one server. Much of the bulk is mechanical batch duplication (a batch_* twin for most single-locus tools) that batch_locus_call already generalizes, plus multiple overlapping synthesis/dossier tools, so many entries do not earn a distinct place.
Coverage of the plant-genomics domain is genuinely broad: locus resolution, sequences, domains, structures, GO/PO/TO enrichment, pathways, homology/orthology, interactions, coexpression, variants, GWAS, literature, and release provenance. Minor gaps remain (no arbitrary-region sequence retrieval, no MSA/phylogenetics, expression limited to BAR eFP and ATTED-II), but agents can work around these.