Loading ToolBench

Reading the current corpus…

FlowRAG MCP server · C · ToolBench

← Browse

FlowRAG

github.com/Zweer/FlowRAG

MCP server for FlowRAG - expose your knowledge base to AI assistants via Model Context Protocol

<=2025-11-257 tools

60/100

Definition Quality72

Protocol Readiness45

Supportability49

Spec posture

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

Static source inference · medium confidence · detected: stateful session

Deprecated protocol patterns detected

Summary

FlowRAG MCP exposes 7 tools with a consistent verb_noun naming pattern (flowrag_*). All tools have descriptions and visible input schemas. However, several tools lack detailed parameter constraints, error handling guidance is missing, output schemas are not documented, and there are no tool annotations (readOnlyHint/destructiveHint). The server targets a knowledge graph RAG domain and implements readable/write operations with appropriate risk marking, but falls short of production-grade quality in parameter validation, output documentation, and error recovery patterns.

Tools (7)

flowrag_entitiesread onlysource verified65/100

List or filter entities in the knowledge graph

flowrag_indexwritesource verified68/100

Index documents into the knowledge base

flowrag_pathread onlysource verified78/100

Find the shortest path between two entities

flowrag_relationsread onlysource verified74/100

Get relations for a specific entity

flowrag_searchread onlysource verified75/100

Search the knowledge base with dual retrieval (vector + graph)

flowrag_search_entitiesread onlysource verified72/100

Semantic search for entities in the knowledge graph by description similarity

flowrag_traceread onlysource verified75/100

Trace data flow upstream or downstream from an entity

Issues (7)

HIGHtool ↗

Output schemas not documented. Tools return structured data but LLMs cannot infer response field names, types, or nesting. Forces agents to guess at downstream usage.

flowrag_indexflowrag_searchflowrag_search_entitiesflowrag_entitiesflowrag_relationsflowrag_traceflowrag_path

HIGHtool-description ↗

Parameter descriptions are minimal (avg 20-30 chars). Lack context on valid ranges, formats, mutual exclusivity, and error conditions. E.g., 'Search mode' doesn't explain when to use hybrid vs naive; 'Max results' lacks min/max bounds.

flowrag_indexflowrag_searchflowrag_search_entitiesflowrag_entitiesflowrag_relationsflowrag_traceflowrag_path

MEDIUMtool ↗

No tool annotations (readOnlyHint, destructiveHint, idempotentHint). While tools are marked READ_ONLY or WRITE in metadata, MCP tool annotations are absent, blocking agent optimization.

flowrag_index

Recommendations

Document output schemas for all tools. Specify field names, types, and nesting. E.g., flowrag_search should return [{id, score, entity_name, summary}, ...]. Use tool.inputSchema pattern with a parallel output schema or include in tool description.
Expand parameter descriptions to 50-150 chars. For 'mode' enum, add: 'Search mode: hybrid (vector + graph), local (graph only), global (vector only), naive (text match). Use hybrid for balanced recall/precision, local for graph-specific queries.' For 'limit', add: 'Max results to return (1-100, default 20).'
Add MCP tool annotations (readOnlyHint, destructiveHint, idempotentHint) to tool definitions in server.ts. Mark flowrag_index as destructiveHint=true (re-indexing may delete old data). Mark search tools as readOnlyHint=true.
Define clear error modes and recovery. E.g., 'If entity not found, return error code 404 with message: "Entity '{entity}' not found. Call flowrag_entities to list available entities." This guides agent retry logic.
Clarify tool selection. Rename 'flowrag_entities' to 'flowrag_list_entities' (list operation) and add explicit description: 'List all entities in the knowledge graph, optionally filtered by type or substring. Use this to discover available entities before search or trace operations.' Rename 'flowrag_search_entities' to 'flowrag_search_entities_by_description' to avoid confusion.
Document pagination. For tools accepting 'limit', add: 'Results are limited to this count. To retrieve more, tools support offset-based pagination (see response metadata). Set limit=20-50 for optimal agent reasoning; larger sets may exhaust context window.'

Spec posture evidence

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

Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained

Score history

Overall score trend

↑ 60 points across a rubric change (v1 → v2)

60/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-22	C	60	<=2025-11-25	v2
2026-03-09	F	0	-	v1