Attach now
Best for ad-hoc diagnosis on a reachable live JVM.
btrace <PID> probe.javaAttach guide →
+ Ask your JVM
+
+ Open-source Java diagnostics
+Ask your JVM
what it’s doing.
+ BTrace helps you safely trace and diagnose live Java applications. Find the slow call, reveal the swallowed exception, and move on—without a restart or redeploy.
+ +Live JVM diagnosticsModern JVM supportZero code changes
+
+
+
+ 01
+ $ btrace -n 'OrderService::* @return if duration>200ms { print method, duration }' 842102
+ ✓ chargeCard306 ms03
+ ✓ processOrder331 ms04
+ watching live methods · ctrl-c to detachBTrace + JFR: generate targeted probes live and add custom events to a continuous recording.
+More context: JFR keeps the history while BTrace adds on-demand instrumentation when questions arise.
+ Explore the workflow +
+
+ Start with a question
+Build your first probe.
+Pick a diagnostic goal. The command is generated in your browser and never leaves your machine.
+
+
+
+
+
+
+ Your BTrace oneliner
Slow method probe
+
+ Need state, multiple hooks, fields, or aggregations?
+That is a reusable Java probe. Start with this equivalent script, then add the advanced feature you need.
+
+
+ Start where it hurts
+Useful answers, not more dashboards.
+BTrace is most useful when you have a concrete question about a running process.
+
+
+ Find a slow call
Trace a method’s duration and keep only the calls that cross your latency budget.
Build a latency probe → +Reveal hidden failures
Capture the stack where a caught or swallowed exception actually escaped.
Build an exception probe → +Inspect database calls
Find slow JDBC statement execution before you start guessing about the database.
Build a SQL probe → +Check for deadlocks
Run a ready-made probe that reports deadlocked Java threads periodically.
View sample probe → +Watch file and socket I/O
Follow files, URLs, and sockets opened by an application you did not write.
Browse samples → +Record structured JFR events
Use custom events when you need low-overhead data for later analysis.
Explore JFR integration → +Grow with the question
From one line to a durable diagnostic.
-
+
- 01
Ask once
Attach an oneliner to answer a focused live question.
+ - 02
Make it repeatable
Turn the oneliner into a Java probe and keep it with your service.
+ - 03
Measure the pattern
Add thread-local state, sampling, aggregations, metrics, or JFR events.
+ - 04
Package the answer
Ship a startup agent or fat agent when the diagnostic belongs in deployment.
+
Fit your environment
Run it where your JVM runs.
Choose the smallest deployment step that answers the question.
+ 01
+ 02
+ 03
+ 04
+
+ Start with an agent
Best when you know the probe before the application starts.
-javaagent:btrace.jarAgent guide →Use a sidecar
Best for interactive troubleshooting in Docker and Kubernetes.
shareProcessNamespace: trueSidecar guide →Build a fat agent
Best when extensions and the agent must travel as one artifact.
./gradlew fatAgentJarFat-agent guide →Targeted by design
Make the next observation a good one.
BTrace is built for safe, temporary instrumentation—not permanent fleet monitoring or application behavior changes.
Keep probes narrow
Prefer exact classes and methods. Use thresholds, sampling, and shallow stacks on hot paths.
Know the attachment boundary
Attach normally requires the same OS user, an attachable JVM, and the right container permissions.
Choose the right recorder
Read the decision guide →Use JFR for continuous, low-overhead history; use BTrace for targeted custom instrumentation.
New diagnostic workflows
Give an AI assistant a careful way to inspect a JVM.
The BTrace MCP server lets compatible AI clients deploy probes and surface results through a diagnostic conversation. BTrace also ships reusable agent plugins that connect BTrace, JFR, and async-profiler into evidence-driven performance workflows.
Which methods in my service take more than 100 ms?
BTrace attaching a bounded latency probe…
Found 2 methods
PaymentClient.authorize · p95 243 ms
OrderService.checkout · p95 271 ms
The BTrace ecosystem
Tools for knowing your JVM.
Focused open-source projects for Java developers and operators.
+
+ See every BTrace project on GitHub →
+ ⌁Flagship
BTrace
Production-safe dynamic tracing and diagnostics for live Java applications.
- Oneliners and Java probes
- Attach or startup-agent modes
⌘Add-ons
BTrace extensions
Curated integrations and richer diagnostic capabilities for BTrace.
Open project →↗Experimental
Jafar
An experimental Java Flight Recorder parser for exploring JVM event data.
Open project →✦AI workflows
Agent plugins
Reusable skills for Claude Code, Codex, and Pi that connect live tracing with JFR and async-profiler analysis.
Open project →