MCP server for EVE (Earth Virtual Expert) - provides tools for querying Earth Observation data, managing document collections, and analyzing Google Earth Engine Python scripts for factuality issues
EVE MCP has 5 tools with basic descriptions and some parameter documentation, but significant gaps in schema completeness, parameter descriptions, and error handling. Tool names follow verb_noun convention (list_, check_, query_, extract_, assess_). However, 3 of 5 tools lack complete input schema visibility in source, and descriptions vary widely in quality. No error handling guidance, no output schemas documented, and no tool annotations. The query_eve tool has reasonable description (~180 chars) but others are sparse. Parameters like 'collections' and 'k' have defaults but lack constraint documentation (e.g., valid collection names, k range). extract_factuality_issues and assess_factuality_issue have verbose descriptions but no parameter-level guidance.
Makes an assessment of a factuality issue identified in a Google Earth Engine Python that attemps to answer an Earth Observation Question. The answer includes a textual assessment, as well as possibly suggestions to update the Google Earth Engine Python code to improve the reliability of the answer
Check the EVE API health status (unauthenticated).
Analyze a Google Earth Engine Python script and extract what aspects or issues are making scientific or data assumptions either explicitly or implicitly and might require factual verification. The extract aspects or issues would be presented to an Earth Observation expert for verification to increase the trustworthiness of the GEE Python code and its results. The response is a a list of json structures, each one describing a specific aspect identified and containing the following fields: 'title', 'description', 'facts', and 'question_for_expert'.
List publicly available EVE document collections.
Query the EVE RAG system and return the assembled answer. Creates a temporary conversation, streams the full response, then deletes the conversation to avoid state accumulation.
Input schemas incomplete or not visible for 3 of 5 tools. extract_factuality_issues and assess_factuality_issue show parameter names and types in docstrings but no explicit JSON Schema registration visible in source. list_eve_collections and check_eve_health have empty input schemas ({}), which is correct but offers no guidance.
Parameter descriptions missing or incomplete. 'collections' parameter in query_eve has a description but no constraint on valid values (e.g., enum of available collections). 'k' parameter lacks range documentation (min/max). extract_factuality_issues and assess_factuality_issue parameters lack descriptions in the visible schema.
No output schemas documented. Tools return JSON strings but no schema is provided describing the structure (e.g., query_eve returns 'answer', 'sources', 'conversation_id' per docstring, but this is not formalized). LLMs cannot plan downstream calls without knowing response structure.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | <=2025-11-25 | v2 |
No error handling guidance. Tools do not document what errors can occur, whether they are retryable, or what the LLM should do next. E.g., query_eve may fail if collections don't exist or EVE API is down, but no recovery guidance is provided.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All tools are read-only (per Risk field), but this is not declared in the tool definition. Modern MCP clients use annotations to optimize caching and retry logic.