From target to evidence.

Getting started

1. Create a project and add your target URL or mobile build.
2. Prove control through the supplied HTTP or DNS challenge. Authorize a bounded workload.
3. Create a scenario: a sequence of actions with explicit success checks.
4. Qualify it with one user. Review the evidence before scaling it.
5. Estimate the run, confirm the spending cap and start. Save a completed run as a baseline.

The private owner login supports setup before WorkOS is configured. Customer authentication requires configured WorkOS credentials and redirect URLs.

Testing modes

Realistic users

Repeat normal journeys with pauses and individual session state.

Gradual load

Ramp concurrent users or set an arrival rate while observing latency and failures.

Release checks

Run saved success criteria and compare with an explicit baseline.

Unusual actions

Explore bounded alternative navigation and inputs within an approved target.

Journey actions

Browser journeys support navigation, click, input, wait, assertions, refresh and back. Protocol journeys support HTTP, server-sent events and WebSockets. Actions use a restricted declarative schema; uploaded scripts are never executed.

Use {{variable}} to reference captured variables. Account and correlation values belong to each user session. Browser selectors should use stable accessible names or test attributes.

Workload controls

Concurrency is the requested number of simultaneous users. Arrival rate expresses new users per second. Duration, ramp, think time, p95 latency limits and error thresholds are recorded with the run. Engine capacity limits can reject a request before it starts.

Protocol users, browser sessions, mobile devices and AI calls have different costs and are counted separately. A thousand protocol users does not mean a thousand browsers.

Set explicit pause boundaries for repeatable randomized user timing. In Advanced JSON, taskMix supplies named task journeys and relative weights. Each iteration chooses one task using the run seed; hybrid tasks also need their own browserSteps.

Break mode uses actions and outcomes you author and review. The browser editor provides double-click, refresh, back and empty-input presets. Replace sample selectors and assert the expected safe rejection or preserved result before estimating a run.

Evidence and baselines

Checks are passed, failed or untested. An interrupted step does not become a pass. Issue fingerprints group repeated findings. Screenshots and traces are downloaded through tenant-authorized links.

A baseline is deliberately promoted from a completed run. Comparisons show changes in checks, failure rate and latency. Export JSON, CSV or PDF for release reviews.

Mobile testing

Upload an Android APK or device-compatible iOS IPA. The native engine uses AWS Device Farm remote Appium sessions. Device admission follows the configured account quota. Missing AWS configuration blocks native runs with a setup reason.

This deployment supports cloud real devices. Local Android emulators and Mac-hosted iOS .app simulators require separate future worker adapters. APK and IPA execution remains unqualified until your AWS account, device access and a real native journey have been tested.

Collector

Create a collector enrollment in your workspace. Install the provided Linux service or Docker package with the one-time enrollment token. The collector reports bounded host metrics and optional read-only PostgreSQL metrics, without granting remote shell access.

Revoke enrollment from the workspace to stop accepting its data.

API and CI

Create a scoped API key in Settings and protect it in your CI secret store. The API uses the /v1 prefix, JSON envelopes and idempotency keys for safe retries.

First request an estimate from POST /v1/runs/estimate with the scenario and workload. Review its credits, then pass its estimateId when starting. Retry an unchanged start with the same idempotency key; a changed workload or price requires a new estimate.

POST /v1/runs
Authorization: Bearer <scoped-api-key>
Idempotency-Key: <unique-release-id>
Content-Type: application/json

{"scenarioId":"your-scenario-id","estimateId":"reviewed-estimate-id","spendingCap":100}

Use the generated OpenAPI contract for exact request fields. Configure completion webhooks in Settings. SSE run events are available at /v1/runs/:id/events.