Loading ToolBench

Reading the current corpus…

AzureAISearchMCP MCP server · C · ToolBench

AzureAISearchMCP

github.com/HeyJiqingCode/AzureAISearchMCP

MCP server for Azure AI Search providing tools for semantic search, vector search, hybrid search, and agentic retrieval capabilities

<=2025-11-25STREAMABLE_HTTP5 tools

62/100

Definition Quality72

Protocol Readiness69

Supportability40

Spec posture

Inferred effective spec <=2025-11-25·current baseline 2026-07-28

Static source inference · medium confidence · detected: Logging

Deprecated protocol patterns detected

Summary

Server provides 5 tools with complete input schemas and descriptions. All tools are read-only search/retrieval operations with well-structured parameters. However, descriptions are generic and lack LLM-optimized guidance on when to use each tool vs. alternatives. Output schemas are not documented. Parameter descriptions exist but lack format constraints, ranges, and dependency hints. Error handling is present but not recovery-focused. No tool annotations (readOnlyHint, etc.) despite all being read-only.

Tools (5)

agentic_retrievalread onlyauthsource verified65/100

Perform agentic retrieval using Azure AI Search knowledge base with reasoning and multi-turn conversation support

hybrid_searchread onlyauthsource verified73/100

Search an Azure AI Search index using hybrid search combining text and vector search with optional filters and facets

searchread onlyauth50/100

Search an Azure AI Search index using simple text search with optional filters, facets, and sorting

semantic_searchread onlyauthsource verified75/100

Search an Azure AI Search index using semantic ranking with optional filters, facets, and sorting

vector_searchread onlyauthsource verified73/100

Search an Azure AI Search index using vector embeddings with optional filters and facets

Issues (5)

HIGHtool ↗

Output schemas not documented. LLMs cannot plan downstream operations or extract required fields from search results. No documentation of what fields are returned, pagination structure, or how to chain results to other tools.

searchsemantic_searchvector_searchhybrid_searchagentic_retrieval

HIGHtool-description ↗

Tool descriptions lack LLM-optimized guidance on WHEN to use each search variant. 'Search using semantic ranking' does not explain when semantic_search is preferable to hybrid_search or vector_search. Agents cannot disambiguate.

searchsemantic_searchvector_searchhybrid_search

HIGHconstrained-input ↗

Parameter descriptions lack format constraints and ranges. 'top' parameter has no min/max bounds (could accept 0 or 1000000). 'vector' array has no element count or value range. 'reasoning_effort' enum values not listed in description.

searchsemantic_searchvector_searchhybrid_search

Recommendations

Document output schema for each tool. Example: 'Returns array of {id, title, score, highlights, facets}. Includes total_count for pagination and next_cursor if more results available.'
Rewrite tool descriptions to disambiguate. E.g., 'Use semantic_search when query intent matters more than exact keywords (e.g., "how to fix a bug"). Use vector_search for similarity matching on embeddings. Use hybrid_search to combine both.' Add 'When to use' section.
Add min/max constraints to numeric parameters in descriptions. E.g., 'top: number of results (1-1000, default 10)'. Add enum values to reasoning_effort: 'minimal|medium|high'.
Apply readOnlyHint annotation to all 5 tools via FastMCP decorator to signal safety. Example: @mcp.tool(name='search', annotations={'readOnlyHint': True})
Enhance error messages with recovery guidance. E.g., 'Endpoint not configured. Set AZURE_AI_SEARCH_ENDPOINT env var or pass endpoint parameter.' instead of bare RuntimeError.

Spec posture evidence

Inferred effective spec: <=2025-11-25.

Relies on Logging (deprecated) - log to stderr or use OpenTelemetry

Score history

Overall score trend

First recorded score · v2 rubric

62/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	C	62	<=2025-11-25	v2