AI Decision System
Replayable, offline-verifiable records for AI-assisted go/no-go decisions: experiment promotions, typed model answers and human-approved actions.
Open JFR recordings or heap dumps and view flamegraphs and JVM dashboards
Translation pending
AI summary · README
Developer toolsJava · VueGitHub @petrboudaSelf-installOwner unconfirmed
Paste into Claude or ChatGPT for installation instructions suited to your environment.
Have you tried it? One tap is enough. No sign-in needed.
petrbouda/jeffrey
Translation pending
Taken from the repository README
Loading the full README. The saved excerpt is shown below.
Open a JFR recording or heap dump and read flamegraphs that finally render fast.
A self-hosted performance analyst for the JVM — flamegraphs, heap dumps, JVM dashboards, and an MCP server for your coding agent, all in one app.
Requires Java 25+. Pre-loaded with sample recordings — nothing to configure, just look around:
Then open http://localhost:8585, go to Recordings → upload a JFR or Heap dump → Analyze,
and Jeffrey builds a profile you can explore. That's it. 🎉
The complete toolbox — Microscope and Hub are the core, the rest round it out:Read the full README ↗Requires Java 25+. Pre-loaded with sample recordings — nothing to configure, just look around:
# with pre-loaded example recordings
docker run -it --network host petrbouda/microscope-examples
# clean, bring your own JFR files
docker run -it --network host petrbouda/microscope
# grab the latest release
curl -L -o microscope.jar \
https://github.com/petrbouda/jeffrey/releases/latest/download/microscope.jar
java -jar microscope.jar
Then open http://localhost:8585, go to Recordings → upload a JFR or Heap dump → Analyze, and Jeffrey builds a profile you can explore. That's it. 🎉
The complete toolbox — Microscope and Hub are the core, the rest round it out:
| Product | What it does |
|---|---|
| Jeffrey Microscope | The standalone analyzer. Open a JFR recording or heap dump — or pull one from the recording volume on an Azure file share — and read flamegraphs, sub-second timelines and JVM & tech dashboards. Runs as a JAR or container. |
| Jeffrey Hub | A containerised maintenance service for Kubernetes. Runs on the shared recording volume your services write their JFR recordings and artifacts to, and keeps it in shape: retention, stale-instance cleanup and LZ4 compression, with a Storage and a Scheduler page. |
| IntelliJ Plugin | Jump from any flamegraph frame straight to the source line in your open IntelliJ — Java & Kotlin — or pull inline source back into the profile. |
| Microscope MCP | Analyse your profiles from an interactive coding-agent session in your own repository — list recordings, query their tables, and pull flamegraph, trace and heap-dump exports, so the profile and the source code sit in front of the same reader. Works with Claude Code (on its v2 MCP runtime) and Codex (v0.147.0+ with the mcp_2026_07_28 feature flag) — one /plugin install microscope@jeffrey or codex plugin marketplace add petrbouda/jeffrey — and with Gemini CLI once it speaks MCP 2026-07-28. |
| Provisioner | One HOCON file lays out your workspace / project / session tree on the shared volume, writes its session markers, and generates the JVM argfile that starts your app under the profiler. |
| Jeffrey JIB | A Jib (Gradle/Maven) extension that wraps your container entrypoint so profiling starts before your app does — no command override, no binaries baked into the image. |
| Jeffrey Events | A lightweight custom-JFR-event library (HTTP, gRPC, database, connection-pool) that powers the tech dashboards. |
Jeffrey is released under the Apache License 2.0 (Apache-2.0).