MCP server for searching European Nucleotide Archive (ENA) datasets. Find RNA-seq studies, retrieve metadata, and discover related publications to validate research hypotheses.
Scoring was not performed
Output schemas not documented. LLMs cannot predict response structure or plan downstream operations. Tools return 'study details' or 'available fields' but the response format (fields, nesting, field types) is invisible to the agent.
Parameter descriptions lack actionable constraints. Many parameters marked 'Optional' without explaining: (a) what happens if omitted, (b) interaction with other params, (c) default behavior. E.g., 'tissue_type' is optional but the description doesn't clarify whether omitting it searches all tissues or returns no results.
No error recovery guidance. Tools (especially find_studies_by_publication) document API limitations and errors but do not guide the LLM on alternatives or workarounds. E.g., 'This tool has known API limitations and will return an error with alternative instructions' tells the user something failed but not what to do next.
Parameter descriptions contain example values in prose. LLMs tend to reuse example values literally (e.g., copying 'PRJDB2345' or 'SRP417965') instead of adapting them. Examples should be moved to formal constraints (enums, patterns) or omitted entirely from descriptions.
build_custom_query accepts 'operator' (=, >=, <=, !=, contains) but the description does not validate which operators work with which field types. E.g., can you use '>=' on a text field like 'organism'? The parameter lacks type-specific constraints.
Tool descriptions reference ENA-specific concepts (e.g., 'tax_id', 'library_strategy', 'SRP/ERP/DRP' format) without defining them for an LLM unfamiliar with bioinformatics. LLMs need definitions to reason about when to call these tools.
generate_download_script is marked WRITE risk but lacks a dry-run or confirmation mechanism. An LLM could accidentally generate and save a malformed script. No pattern:confirmation-request support is visible.
Pagination not documented. search_rna_studies and search_studies_by_keywords accept 'limit' but do not document how to retrieve the next page of results (no offset, cursor, or next_token guidance).
list_library_types and get_result_types have minimal descriptions ('List all available...' and 'Get all available...'). Descriptions are 60-70 chars when best practice is 50-200 chars with explicit guidance on WHEN and WHY to call them.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 0 | 2026-07-28+ | v2 |
| 2026-03-09 | C | 61 | - | v1 |