Spring Cloud code knowledge graph and MCP server for AI coding agents. Provides read-only knowledge graph access to Spring Boot projects including HTTP endpoints, Feign clients, MyBatis mappers, runtime configuration, and inter-service call flows.
Springgraph provides a well-intentioned MCP server for Spring Cloud code analysis with 4 tools addressing real agent needs (code navigation, impact analysis, flow tracing). However, the implementation shows significant gaps in definition quality. All tools have basic descriptions (60-120 chars), which meet the minimum bar but lack the specificity recommended for LLM tool selection. Input schemas are present and typed, but parameter descriptions are sparse and often generic. Output schemas are not documented in the provided source, forcing agents to infer result structures. Error handling is listed as 'true' but no recovery guidance is evident in tool descriptions. The tools themselves are well-named with clear action verbs (find, get, trace, analyze) and logical single responsibilities, which is the primary strength. However, the lack of documented output structures, missing parameter constraints (enums, ranges), and absence of error recovery guidance place this in the 'fair' category (C-range) rather than 'good' (B-range).
Get an overview of all Spring assets in the project. Returns counts of controllers, services, mappers, feign clients, etc. Survey of services, middlewares, and sensitive config properties. Sensitive values are never returned — only key, file:line, and a sensitivity flag.
Locate entry points by URL, controller class, Feign name, MQ topic, or scheduled task. Returns the endpoint symbol, handler file:line, and call-chain head.
Analyze the impact of a Spring method. Returns callers, callees, endpoints, transaction boundaries, and related exception handling. Analyze a method's callers, callees, endpoints, transaction boundary, exception handlers, and downstream SQL operations.
Trace the flow from an endpoint to its dependencies. Shows the call chain from controller to service to mapper/feign. Trace the call flow from an HTTP endpoint, controller, service, repository, mapper, SQL, and table. Depth 1-3 returns 2-4 levels; depth 5 returns the full Endpoint→Controller→Service→Mapper→SQL→Table chain.
Output schemas not documented in source code. Tools describe what they return in natural language (e.g., 'Returns the endpoint symbol, handler file:line, and call-chain head') but no structured output schema is visible. Agents cannot plan downstream tool calls or extract typed data reliably without documented result schemas.
Parameter descriptions are minimal or missing constraint details. Example: spring_find_entry 'query' parameter has description 'Optional search query to filter by controller name or route path' but lacks format guidance (regex, length limits, case sensitivity). spring_trace_flow 'depth' parameter defaults to 5 with range 1-3 or 1-5 (unclear from description). Agents cannot validate inputs without explicit constraints.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
No error recovery guidance in tool descriptions. errorReporting is listed as 'true' but tool descriptions do not explain what errors are possible, when to retry, or what to do if a lookup fails (e.g., 'If spring_find_entry returns no results, try a broader query or list all endpoints via spring_assets_overview'). Agents are left to guess on failure.
spring_assets_overview has empty input schema (no properties). Tool description mentions 'Returns counts of controllers, services, mappers, feign clients' but does not explain if there are filters (e.g., by package name, sensitivity level) or if it always returns the full overview. Empty schema suggests no parameters, but the interface could be clearer.
Parameter naming inconsistencies may confuse LLMs. spring_find_entry uses both 'query' and 'url' to specify search targets (query is 'Optional search query to filter by controller name or route path'; url is 'Optional endpoint path alias for query'). The distinction between these two parameters is unclear, are they mutually exclusive, additive, or interchangeable?
Tool descriptions do not clearly distinguish use cases between similar tools. spring_trace_flow and spring_method_impact both analyze code relationships. When should an agent call one vs. the other? The descriptions overlap ('trace flow from endpoint to dependencies' vs. 'analyze impact of a method'), increasing risk of tool misselection.