What is Atlas?

Atlas runs JMeter load tests across a fleet of machines and shows you what happened while it is still happening.

You bring a JMeter test plan. Atlas takes care of the parts around it: getting the plan and its data onto every load generator, starting them together, collecting results as they arrive, and shutting the machines down when the test finishes.

What it does

Runs your plan on many machines at once. One machine can only generate so much load. Atlas copies your plan to as many load generators as you ask for and runs them in parallel, so the load adds up.

Divides your test data. If your plan reads from a CSV, each generator gets its own slice of the rows rather than every machine replaying the same records.

Shows results live. Response times, throughput, active users and error rates appear while the test runs, broken down per request and per thread group.

Starts and stops machines for you. Load generators that are switched off are powered on when a test needs them and switched off again when it ends, so you are not paying for machines between runs.

Three ways to use it

Everything can be done from the web interface or from the command line — they talk to the same system and follow the same permissions.

Web interface	Day-to-day work: upload files, start a test, watch it run, read the report. Start at Getting Started.
Command line	Automation: run a test from a CI pipeline and fail the build when it regresses. Start at CLI Installation.
An AI assistant	Describe the test in plain language and let Claude Code, Copilot or Cursor drive the CLI for you. Start at Atlas for AI Agents.

What you need

• A JMeter test plan (.jmx) and any data files it reads. If you don't have one yet, the recorder can produce one by watching you click through your application.
• An Atlas account. Your administrator creates this and gives you access to the projects you should see.
• Load generators. Most teams share a pool that is already set up. If you need to add your own machines, see Load Generators.

How a test run works

1. You pick a project, a test plan, and how many load generators to use.
2. Atlas reserves those machines so no other test can use them at the same time, and powers on any that are switched off.
3. Your plan and its data files are copied to each machine, with data files divided between them.
4. JMeter starts on every machine at once.
5. Results stream in while the test runs.
6. When the test ends, the machines are released and switched off.

TIP

A run is identified by a test ID that looks like SHOP-20260818-194201 — the project name, the date, and the time it started. You will see it in the dashboard URL, and the CLI prints it when a run begins.