This Elasticsearch MCP server has 10 tools with basic schemas and descriptions, but falls short of production quality in multiple critical areas. All tools have descriptions (10-60 chars, well under the 194-char baseline), and input schemas are present with type definitions. However, descriptions are terse and lack actionable context for LLM selection. Parameter descriptions exist but are minimal (avg ~20 chars vs 72-char baseline). No parameter type constraints (enums) for categorical inputs. No output schemas documented. Error handling is absent from the code, no guidance for LLM recovery, no categorization of error types, and no validation error messages. Security: credentials are environment-injected (good), but no per-tool permission declarations or audit logging. Tools follow verb_noun naming but lack composition clarity, no documented dependencies between tools, no chaining field inclusion in outputs. Missing pattern adoption: no tool annotations (readOnlyHint/destructiveHint/idempotentHint), no MRTR for confirmation on destructive ops, no per-request logLevel. Overall, the server is functional but under-documented for agent use.
Get cluster information
Create a new index
Delete a document
Delete an index
Get a document by ID
Check Elasticsearch cluster health
Index a document
Descriptions are terse (10-60 chars) and lack actionable context. Tool descriptions must state WHAT the tool does, WHEN to use it, and any prerequisites. Current examples: 'Check Elasticsearch cluster health' (39 chars) and 'Delete a document' (17 chars) do not explain when to call them vs related tools, what they return, or what errors to expect.
No parameter type constraints or enums. Parameters like those in es_create_index's 'mappings' and 'settings' (type: object) lack enum, pattern, or minLength/maxLength constraints. Free-form strings and objects invite hallucinated or invalid values. LLMs cannot distinguish valid configurations from noise.
No output schemas documented. Tools return JSON strings via 'text' content type, but the structure and fields are not declared. LLMs cannot plan downstream calls or extract data reliably without knowing response shape.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 48 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 35 | - | v1 |
List all indices in the cluster
Search documents
Update a document
No error handling or recovery guidance in tool definitions. The code has no explicit error responses, validation messages, or categorization (retryable vs fatal). Agents hitting errors receive only raw exceptions with no actionable next steps.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Destructive tools (es_delete_index, es_delete_document) and write tools (es_create_index, es_index_document) lack semantic hints, making it impossible for agents to reason about reversibility or side effects.
Destructive operations (es_delete_index, es_delete_document) lack confirmation mechanism. No dry-run mode or multi-round-trip request (MRTR) pattern to prevent accidental deletion. Agents should be forced to confirm before irreversible actions.
No per-tool permission declarations or audit logging. Tools lack scope declarations (e.g., 'read:indices', 'write:documents', 'admin:cluster') and there is no audit trail of who called what and when. Compliance and security traceability are missing.
No result pagination or limits documented. es_list_indices and es_search could return hundreds or thousands of results, bloating the context window. No mention of page, limit, offset, or total_count in schemas. LLMs cannot plan for large result sets.
Parameter descriptions are minimal (avg ~20 chars). E.g., 'Index pattern (default: \'*\')' lacks explanation of what patterns are accepted, syntax rules, or common use cases. Parameter descriptions should average 72+ chars and include format, range, and constraints.
No tool chaining documentation. When es_create_index succeeds, what fields does the response include that downstream tools (e.g., es_index_document) accept? No cross-tool field naming consistency declared (e.g., does 'index' always mean the same thing?).