SmartBlitz AI for JMeter · three free, open-source plugins

Every release, re-scripted automatically. Zero manual effort.

When the application changes, the toolkit re-records the journey with Playwright, finds every dynamic value, applies your saved correlation templates, and hands back a working JMeter script. No re-recording by hand, no regex hunting.

I’m SmartBlitz. I re-script it so you don’t have to.
Every release
▸ Re-record journeys (Playwright) ✓
▸ Correlate dynamic values ✓
▸ Apply saved templates ✓
▸ Run and watch live ✓
Manual rescripting steps: 0
0manual rescripting steps per release
3plugins, each installable on its own
Liveper-transaction results while the test runs
Apache 2.0open source, no account, no telemetry
Release day, before and after

From a day of rescripting to nothing to do

Dynamic values change with every build: session tokens, CSRF tokens, IDs. Fixing a script after each release is slow, repetitive work. The toolkit does it for you, and learns from every correlation you approve.

Manual scripting

Repeated for every script, every release
  1. Re-record the journey by hand in a browser or proxy
  2. Replay it and read the failures to find what changed
  3. Find each token and ID that moves between runs
  4. Write a regex extractor for every one of them
  5. Wire each variable into the right requests
  6. Debug the replay, then do it again for the next script
Hours of repetitive, error-prone work per release.

With SmartBlitz AI

Hands-off after the first review
  1. A Playwright script replays the journey in a headless browser
  2. Compare Scripts shows exactly what changed since last release
  3. The Auto-Correlator finds dynamic values automatically
  4. Saved templates apply the right extractors and variables
  5. Review a new value once; it is reused on every later release
  6. The script runs and streams into the Insight Dashboard
Review once, reuse every release. No manual rescripting.
Architecture

How the automation fits together

A Playwright-driven pipeline produces and repairs the script. Saved templates carry the knowledge forward. The MCP Bridge and the Insight Dashboard sit around it as optional, independent plugins.

Animated architecture: a Playwright script is replayed in a headless browser, the traffic becomes a JMeter script, the Auto-Correlator finds dynamic values and applies saved templates, the test runs locally or remotely, and results stream into the Insight Dashboard. The MCP Bridge exposes the open test plan to AI agents.
Automated release pipeline AI agent access (MCP) Live results Supporting tools
01

Record with Playwright

A Playwright script drives a real browser through the journey, so recording is code you can re-run, not a manual session.

02

Build the JMeter script

The captured traffic becomes a JMeter test plan. Compare Scripts diffs it against the previous version.

03

Correlate automatically

The Auto-Correlator finds session tokens, CSRF tokens and IDs, and applies saved templates from the template store.

04

Run and observe

Run locally or on a remote host over SSH, and watch results arrive live in the Insight Dashboard.

Playwright implementation

Recording is code, so it can run every release

Because the journey is a Playwright script, re-capturing it after a release is a re-run, not a rewrite. Three ways to get a script, depending on where you start.

Start fresh

Generate Playwright Script

Create a new Playwright script for a journey and turn the recorded session into a JMeter script.

After a release

Re-record Existing Script

Re-run the stored journey against the new build and rebuild the JMeter script from the fresh traffic.

Explore

Record Interactively

Drive the browser step by step yourself and capture exactly what you did.

Real browser, real experience. The same browser tooling powers Check Page Load Timing Under Load, which measures how long a real browser takes to render a page while your load test runs. It answers the user-experience question that server response time alone cannot.
The toolkit

Three plugins, each stands on its own

Install only the ones you need.

JMeter Workbench

Correlation, scenario recording, script diffing, page-load timing under load and test data generation. One toolkit for keeping scripts working as the application changes.

  • —Correlate a Script: finds dynamic values (session tokens, CSRF tokens) and suggests a regex for each, reviewed before anything is applied
  • —Record a New Scenario: from a fresh Playwright script, a re-recorded script, or a live interactive browser
  • —Compare two script or recording versions to see exactly what changed
  • —Check Page Load Timing Under Load: real browser rendering time, not just server response time
  • —Test Data Generator: names, emails, sequential IDs and weighted distributions for a CSV Data Set Config
View on GitHub →

MCP Bridge

A localhost-only MCP server that exposes your live, currently open test plan to any MCP-capable AI agent. No exporting or reimporting a .jmx file.

  • —Read and edit the actual open test-plan tree from Claude Code, Codex or opencode
  • —A built-in Chat tab: talk to Claude Code about your test plan without leaving JMeter
  • —Deletable conversation history, with a confirmation before anything is discarded
  • —Split out from Workbench so it installs on its own, with no dependency on it
View on GitHub →

Insight Dashboard (AI)

A real-time, per-transaction results listener. Watch a test while it is running, not after exporting a .jtl and opening something else.

  • —Live stat cards and response-time, throughput and error-rate charts, updating as the test runs
  • —An expandable tree table per transaction, with SLA threshold highlighting and P90/P95/P99
  • —Request and Response Data split into Headers, Body and Cookies, each independently searchable
  • —CSV/Excel export, plus an optional AI Results Summary and Chat (local Gemini CLI, no network calls otherwise)
View on GitHub →
Key features

Six things that do the heavy lifting

Real screenshots from JMeter. Hostnames, file paths and credentials are blurred. Click any image to enlarge.

01 · Insight Dashboard

Watch the test live, then ask why

Live charts, a per-transaction table with P90, and drill-down to the last sample's request and response.

  • Live response time, throughput, error rate and response-code charts
  • Per-request stats grouped by transaction, filter and sort while the test runs
  • AI Results Summary and AI Chat on top of the run, using your local Gemini CLI
02 · Correlation

Dynamic values found and fixed for you

The Auto-Correlator reads a recording, lists every value produced by one response and reused later, and applies the extractor.

  • Analyze a HAR or JTL, review candidates, apply to a new .jmx (original untouched)
  • Save extractors as reusable templates, so next release is one click
  • Test Regex before applying, optional fail-loudly assertions
03 · Automatic script recording

Record once, regenerate every release

Generate a Playwright script, re-record an existing one, or record interactively. Each path ends in a HAR and a .jmx.

  • Playwright drives a real browser through the journey
  • Re-record with the scenario template to refresh the script after a UI change
  • Mark transactions live while you click through
04 · Script comparison

See exactly what the release changed

Compare last release's script against the fresh recording: added, removed and changed requests, with field-level before and after.

  • Works across .jmx, .har and .jtl
  • Changed requests show category, name, old and new value
  • Review a short diff instead of re-reading the whole script
05 · Test Data Generator

Realistic data without a spreadsheet

Define columns, pick a type, preview ten rows, generate a CSV for the run.

  • Names, emails, phones, dates, UUIDs, sequences, numbers, lists, weighted values
  • Patterns and templates, null percentage per column
  • Instant preview before generating
06 · UI render under load

Measure how fast the page renders, under load

Capture Render Timing starts your load test, then replays the scenario once in a real browser while the load is active. You get paint and load timings per transaction, plus a per-request network breakdown.

  • DOMContentLoaded, Load, First Paint, First and Largest Contentful Paint per transaction
  • Network timings per request: DNS, connect, SSL, TTFB, download
  • One real browser on top of real load, not many browsers pretending to be users
More in the app

Everything in one workbench

WorkbenchOne launcher for scripting, execution and data tools.
MCP BridgeLets an AI agent work on the test plan open in JMeter, with undo as a safety net.
AI Results SummaryFindings, likely causes and recommendations from the run.
Remote runStart a test on a remote host over SSH and feed the live dashboard.
Time saved

Estimate the effort you get back

Enter numbers for your own suite. Starting values: 6 clients under baseline testing, 5 scripts each, 5 hours per script to rescript by hand, 3 releases a year.

0 h
saved per release
0 h
saved per year
0 days
of 8-hour working days
Get started

Install in a minute

  • 1Pick the plugin or plugins you need from GitHub.
  • 2Drop the jar in lib/ext.
  • 3Restart JMeter. No account, no telemetry.