Service Dependency Intelligence for Contract-Aware Agent Design in Enterprise Microservice Architectures

LLM-based coding agents designing solutions inenterprise microservice architectures must understand servicedependencies, API contracts, and event topologies to producecorrect, deployment-safe designs. We present a service de-pendency intelligence system that constructs and maintains aqueryable dependency graph from static code and configurationanalysis, enabling agents to perform contract-aware design. Thesystem extracts over 3,200 dependency edges across 74 distinctdetection signal types from approximately 900 repositories ata Fortune 50 retailer, queryable in under 500ms via a RediS-earch index. Integrated into the agent’s design workflow as amandatory pre-modification step, the system enforces contract-aware architecture rules: agents must enumerate all affectedconsumers, specify backward-compatible migration strategies,and define deployment ordering before proposing any API orevent schema change. The dependency graph achieves 100%precision on detected edges (team-validated, zero false positives)with a 5–10% recall gap from scan depth limitations. Withover 1,800 total interactions across a conversational AI assistant,catalog browser, and automated design pipelines, the systemdemonstrates that static dependency intelligence—without run-time instrumentation—provides sufficient context for agents toproduce designs that respect the enterprise service topology.