Each tool maps to one Actuator endpoint and takes that endpoint's response as its argument, so the server never needs network access to the service under diagnosis and never needs its credentials. That fits the common case where the Actuator endpoints sit behind a VPN or an operations proxy: you can copy a response out but you cannot expose the port. analyze_env is often the one worth running first, because its findings are about exposure rather than performance — exposed secrets and risky configuration — and it is the one that costs something to leave unrun.
A Spring Boot Actuator response analyser with seven tools, one per endpoint, each taking that endpoint's JSON as its input.
- Health: analyze_health diagnoses the unhealthy components in a /health response and returns recommendations.
- Metrics: analyze_metrics detects JVM heap pressure, non-heap metaspace pressure, high error rates, connection pool exhaustion and GC issues.
- Configuration exposure: analyze_env reads an /env response for exposed secrets, risky configurations and missing production settings.
- Application structure: analyze_beans detects circular dependencies, scope mismatches and bean architecture issues, and analyze_startup reads the /startup endpoint available from Spring Boot 3.2 onward for slow bean initialisation, heavy auto-configurations and startup bottlenecks.
- Caches and logging: analyze_caches lists the registered caches and flags unbounded ones as a memory-leak risk, and analyze_loggers detects DEBUG or TRACE levels in production, inconsistent levels across packages and verbose framework logging.
The Actuator endpoint responses themselves as JSON — every tool takes the JSON rather than a URL, so reaching the endpoints is your job. analyze_startup needs the /startup endpoint, which its description places at Spring Boot 3.2 and later. No credentials are configured.
One command — npx -y mcp-spring-boot-actuator