MCP JVM Diagnostics
JVM diagnostics MCP server for thread dump and GC log analysis. Detects deadlocks, lock contention, and GC pressure. Runs via npx on Node.js — no JVM or Docker required.
MCP JVM Diagnostics
A Model Context Protocol (MCP) server that gives AI assistants the ability to analyze JVM thread dumps and GC logs. It detects deadlocks, identifies lock contention hotspots, analyzes garbage collection pressure, and recommends JVM tuning parameters.
Why This Tool?
JVM diagnostic MCP servers exist (TDA, jfr-mcp, Arthas) — but they're all Java-based, requiring a JVM runtime just to diagnose JVM problems. This tool runs on Node.js via npx — no JVM, no Docker, no SSH.
It analyzes offline artifacts (thread dumps, GC logs, heap histograms) rather than requiring a running JVM. Paste a thread dump or GC log and get instant analysis.
Features
- 6 MCP tools for comprehensive JVM diagnostics
- Thread dump analysis — deadlock detection, contention hotspots, thread state breakdown
- GC log analysis — pause statistics, heap trends, memory leak detection
- Heap histogram analysis — memory leak candidates, classloader leaks, finalization issues
- JFR summary — Java Flight Recorder file analysis
- Unified diagnosis — cross-correlates thread and GC data
- Supports G1, ZGC, Parallel, Serial, and Shenandoah GC formats
- No external dependencies — works on local text input, no API keys needed
Pro Tier
Generate exportable diagnostic reports (HTML + PDF) with a Pro license key.
- Full JVM thread dump analysis report with actionable recommendations
- PDF export for sharing with your team
- Priority support
$9.00/month — Get Pro License
Pro license key activates the generate_report MCP tool in mcp-jvm-diagnostics.
Installation
npx mcp-jvm-diagnostics
Or install globally:
npm install -g mcp-jvm-diagnostics
Configuration
Claude Desktop
Add to ~/.claude/claude_desktop_config.json:
{
"mcpServers": {
"jvm-diagnostics": {
"command": "npx",
"args": ["-y", "mcp-jvm-diagnostics"]
}
}
}
No environment variables needed — this tool works on text input you provide.
Quick Demo
Try these prompts in Claude (paste your JVM output):
- "Analyze this thread dump: [paste jstack output]" — Detects deadlocks, contention hotspots, and thread state breakdown
- "Analyze this GC log: [paste GC log]" — Shows pause statistics, heap trends, and tuning recommendations
- "Compare these heap histograms to find memory leaks: BEFORE: [paste] AFTER: [paste]" — Identifies growing classes, classloader leaks, and finalizer issues
Tools
analyze_thread_dump
Parse a JVM thread dump and analyze thread states, detect deadlocks, and identify lock contention.
Parameters:
thread_dump— The full thread dump text (fromjstack <pid>,kill -3, or VisualVM)
Example prompt: "Analyze this thread dump and tell me if there are any deadlocks"
analyze_gc_log
Parse a GC log and analyze garbage collection patterns, pause times, and memory pressure.
Parameters:
gc_log— The GC log text (from-Xlog:gc*or-verbose:gc)
Example prompt: "Analyze this GC log and tell me if there are any performance issues"
analyze_heap_histo
Parse jmap -histo output and detect memory leak candidates, object creation hotspots, and classloader leaks.
Parameters:
histo— The jmap -histo output text (fromjmap -histo <pid>orjmap -histo:live <pid>)
Example prompt: "Analyze this heap histogram and tell me if there are any memory leak candidates"
diagnose_jvm
Unified JVM diagnosis combining thread dump and GC log analysis for comprehensive root cause analysis.
Parameters:
thread_dump(optional) — Thread dump textgc_log(optional) — GC log text
Example prompt: "I have both a thread dump and GC log from the same time — diagnose what's wrong"
compare_heap_histos
Compare two jmap -histo snapshots taken at different times to detect memory growth patterns and leak candidates.
Parameters:
before— The first (earlier) jmap -histo outputafter— The second (later) jmap -histo output
Example prompt: "Compare these two heap histograms and tell me what's growing"
Detects:
- Classes with growing instance/byte counts (leak candidates)
- New classes that appeared between snapshots
- Classes that disappeared (GC'd or unloaded)
- Overall heap growth rate
- Classloader leaks and finalizer queue growth
Collecting JVM Diagnostics
Thread Dump
# Using jstack
jstack <pid> > thread-dump.txt
# Using kill signal (Linux/Mac)
kill -3 <pid>
# Using jcmd
jcmd <pid> Thread.print > thread-dump.txt
GC Log
Add these JVM flags to your application:
# Java 9+ (unified logging)
-Xlog:gc*:file=gc.log:time,level,tags
# Java 8
-verbose:gc -XX:+PrintGCDetails -XX:+PrintGCDateStamps -Xloggc:gc.log
Contributing
- Clone the repo
npm installnpm run buildnpm test
Limitations & Known Issues
- Text input only: Thread dumps, GC logs, and heap histograms must be provided as text. The server cannot attach to a running JVM or capture data automatically.
- Java 9+ GC logs: The GC log parser is optimized for unified logging format (
-Xlog:gc*). Legacy-verbose:gcformat (Java 8) has basic support but may miss some events. - Shenandoah GC: Limited support. G1, ZGC, Parallel, and Serial are fully supported.
- Virtual threads (Java 21+): Thread dump parser handles virtual threads but analysis recommendations are tuned for platform threads.
- Heap histo comparison: Requires standard
jmap -histoorjcmd GC.class_histogramformat. Custom formats or truncated output may not parse correctly. - Deadlock detection: Detects monitor-based deadlocks. ReentrantLock deadlocks may not be detected if lock addresses are not visible in the thread dump.
- No historical analysis: Each analysis is a point-in-time snapshot. For trend analysis, compare multiple snapshots manually.
- HotSpot/OpenJDK only: Parser targets HotSpot/OpenJDK thread dump format. GraalVM native-image or Eclipse OpenJ9 dumps may parse incompletely.
- Classloader leak detection: Heap analysis flags growing ClassLoader instances but cannot definitively prove a leak without memory profiler data.
Part of the MCP Java Backend Suite
- mcp-db-analyzer — PostgreSQL/MySQL/SQLite schema analysis
- mcp-spring-boot-actuator — Spring Boot health, metrics, and bean analysis
- mcp-redis-diagnostics — Redis memory, slowlog, and client diagnostics
- mcp-migration-advisor — Flyway/Liquibase migration risk analysis
License
MIT
相關伺服器
Scout Monitoring MCP
贊助Put performance and error data directly in the hands of your AI assistant.
Alpha Vantage MCP Server
贊助Access financial market data: realtime & historical stock, ETF, options, forex, crypto, commodities, fundamentals, technical indicators, & more
AntBot MCP Server
A TypeScript MCP server for integrating with the AntBot AI-based RPA platform, handling tool listing and execution.
VibeUE
Unreal Engine MCP
Vibe-Coder
A server for a structured, LLM-based coding workflow, from feature clarification and planning to phased development and progress tracking.
SolHunt Solana Wallet Intelligence
Solana wallet health analysis platform and top-notch dev tool. Helps people and agents to recover their SOLs from burner and old wallets super securely. Features a complete trustless recovery flow natively via MCP: preview yields, build unsigned transactions, and sign locally.
Scrnr
Take website screenshots with Scrnr.io
MAVAE - IMAGE TOOLBOX
A creative toolkit for AI agents to generate, edit, and manage images, models, and collections using the MAVAE API.
Pickapicon
Quickly retrieve SVGs using the Iconify API, with no external data files required.
Fused MCP
A Python-based MCP server for data scientists to run Python code with a Claude client.
MCP for Dart
A Dart SDK for building MCP servers and clients.
Arduino MCP Server
Control an Arduino board from your computer using AI commands.