Production-ready MCP server for Reactome pathway and systems biology data
The Reactome MCP server demonstrates solid tool definitions with clear naming conventions, reasonable descriptions, and proper JSON Schema implementations. All 18 tools follow verb-noun naming patterns (search_, get_, find_, etc.) and include descriptions. Input schemas are explicitly defined with types and constraints. However, there are gaps in parameter descriptions, missing output schema documentation, and incomplete error handling guidance. The server excels at constraint specification (enums, min/max bounds) but lacks depth in describing WHEN tools should be used and what downstream tools they chain to. No tool annotations (readOnlyHint, idempotentHint) are present despite all tools being read-only operations.
Perform pathway enrichment analysis on a list of genes or proteins
Compare pathways across different species for orthology and conservation
Find disease-associated pathways and mechanisms
Find all pathways containing a specific gene or protein
Get information about protein complexes and their components
Find pathways and targets for a drug or chemical compound
Find orthologous proteins or genes across different species
Get comprehensive information about a specific pathway
Output schemas not documented. Tools return data from Reactome API but LLMs cannot predict response structure, making it impossible to plan downstream tool chains or extract required fields reliably.
Missing descriptions for optional parameters. 'species' appears in multiple tools (find_pathways_by_gene, get_top_level_pathways, compare_pathways, analyze_pathway_enrichment) but descriptions are minimal or absent, leaving LLMs unclear on format (species name vs NCBI taxon ID).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 51 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Get hierarchical structure and parent/child relationships for a pathway
Get interaction and relationship data within a pathway
Get all molecules (proteins, genes, compounds) participating in a pathway
Get all reactions and events that comprise a pathway
Get protein information and its associated pathways
Get detailed information about a specific reaction or event
Get list of all species available in Reactome
Get top-level (root) pathways for a given species
Search for biological pathways by name, description, or keywords
Validate and get metadata for a Reactome entity ID
No tool annotations despite all 18 tools being read-only operations. Missing readOnlyHint would enable agents to infer safety and retry behavior without explicit documentation.
Error handling guidance missing. Tools do not document expected failure modes (API timeout, invalid ID format, resource not found) or recovery steps. LLM receives raw API errors with no actionable next steps.
Inconsistent parameter naming and descriptions. 'id' parameter used across tools but descriptions vary in specificity. Some mention 'Reactome stable identifier', others just say 'stable identifier'. This inconsistency increases LLM confusion when chaining tools.
No documentation of tool dependencies and chaining. For example, search_pathways can return pathway IDs, which feed into get_pathway_details, but this dependency chain is not documented. Forces agents to infer connections.
No pagination or result limiting guidance in descriptions. Tools like search_pathways and find_pathways_by_disease accept 'size' parameter but descriptions do not explain default behavior or consequences of omitting this parameter.
Vague descriptions on analysis tools. 'analyze_pathway_enrichment' description does not explain what enrichment analysis means in this context, when to use it vs simpler lookup tools, or how to interpret results.