MCP server that provides tools to interact with the UCSC Genome Browser API. Exposes various endpoints for querying genomic data, sequences, tracks, and metadata from the UCSC Genome Browser.
The server defines 10 tools with consistent naming (verb_noun pattern: find_, list_, get_) and complete JSON Schema input definitions. All tools have descriptions (median ~120 chars, within the 10 - 1024 char baseline). However, tool descriptions are generic and lack WHEN-to-use context; parameter descriptions are minimal (10-20 chars on average, well below the 72-char baseline); output schemas are undocumented (no structured response definitions provided in the source); error handling is absent from tool definitions; and no tool annotations (readOnlyHint, idempotentHint) are declared despite all tools being read-only. The server correctly uses enum constraints for parameters like 'browser', 'category', 'status', 'level', and 'format', which reduces hallucination risk. However, parameter descriptions like 'Filter by browser availability' lack actionable detail (e.g., what each enum value means in the UCSC context). The 'max_items' pattern across multiple tools is good (suggests pagination awareness), but no pagination cursor/next_token mechanism is documented. The schema is well-formed JSON Schema, but the source does not reveal response structure, so output documentation cannot be scored.
Search for a genome in the UCSC browser using a search string. Supports advanced search with +word (force inclusion), -word (exclusion), and word* (wildcard).
Retrieve DNA sequence from a specified genome assembly. Can retrieve entire chromosome or specific coordinates.
List chromosomes in an assembly hub, track hub, or UCSC database genome. Optionally filter by specific track.
List download files available for a specified UCSC genome assembly.
List UCSC Genome Browser database genomes from assembly hub host (GenArk). Can also test for existence of a specific genome.
List all genomes available in a specified track or assembly hub.
Output schemas completely undocumented. No tool lists what fields it returns, data types, or structure. LLMs cannot plan downstream calls or extract required fields.
Parameter descriptions are minimal (8-20 chars average) and lack actionable detail. E.g., 'Filter by browser availability' does not explain what mustExist/mayExist/notExist mean in UCSC context; 'Maximum number of items to return' lacks range guidance (is 1000000 realistic? Can it cause timeouts?). LLMs need explicit constraints and use-case guidance.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 54 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
List all available public track hubs in the UCSC Genome Browser.
List the schema (field definitions) for a specified data track.
List all data tracks available in a specified hub or UCSC database genome.
List all UCSC Genome Browser database genomes available on the database host.
Tool descriptions are generic and lack WHEN-to-use context. E.g., 'List all available public track hubs in the UCSC Genome Browser' does not say when an LLM should call this vs list_ucsc_genomes or list_hub_genomes. Unclear tool distinctions force LLMs to guess, increasing misprediction.
No tool annotations (readOnlyHint, idempotentHint) declared. All 10 tools are read-only and safely idempotent, but LLMs cannot verify this from the tool definition. Without annotations, agents treat these as potentially destructive and apply unnecessary confirmation/undo logic.
No error handling guidance or recovery strategies defined. If an API call fails (e.g., invalid genome name, malformed hub URL, timeout), LLMs have no recovery path. Error messages should be documented as part of the tool contract.
Pagination mechanism unclear. Multiple tools accept max_items, but no cursor, next_token, or offset is documented. How do agents paginate large result sets? Do they call the tool again with an offset? Is there a next_cursor in the response?
Parameter 'hub_url' accepts free-form strings with no format validation or examples. LLMs may hallucinate invalid URLs or confuse hub syntax. Should specify URL format (e.g., 'https://...', path structure) in the description.