LoadWave
Distributed load testing with a live dashboard — one binary, correct percentiles across machines, and an exit code your pipeline can act on.
LoadWave generates HTTP load from a fleet of worker processes, merges what they measure into one coherent picture, and shows it live in a browser — or prints a report and exits non-zero, if that is what your pipeline needs.
It is one binary. No agent to deploy alongside it, no database to stand up, no separate UI to host.
loadwave demo # nothing to set up — see it working now
loadwave run test.yaml # run your test, print a report, exit
loadwave run test.yaml --ui # ...and watch it in a browserWhere to start
Install it
One go install, or a binary from Releases.
Quickstart
See the dashboard in about a minute, with nothing to configure.
Your first test
A YAML test with stages, think time and thresholds, explained line by line.
Configuration reference
Every field of the YAML format, with its default and the reason it exists.
The rest of the site
Getting started installs LoadWave and walks a first test end to end.
Guides are task-shaped: choosing between YAML and Go, writing each of them, gating a pipeline on thresholds, reading the dashboard, taking results away, running across machines, and working out what a strange number means.
Reference is exhaustive rather than narrative — the configuration format, the CLI, the metric set, the Go SDK surface and the REST API.
Concepts explains how the coordinator, agents and workers fit together, and how measurements from many machines are merged into numbers that mean something.
What it is not
Worth knowing before you invest time:
HTTP only, for now. Other protocols are on the roadmap. Runs are in-memory — there is no database, and history is lost when the coordinator exits. The control plane is unauthenticated, so it belongs on a trusted network rather than the public internet.
The project is early. The architecture is settled and the whole path —
coordinator, agents, worker processes, merged metrics, dashboard — works and is
covered by tests that spawn real subprocesses. The API may still change before
v1.0.0.