A usage measure, not a guess
The pre-run amount is a safety reserve. It is not a guaranteed quote because different journeys produce different work.
Usage / Matrix Units
Understand the unit behind every evidence-backed browser run.
Matrix Units are the customer-facing usage measure for Matrix QA. A run reserves capacity before it starts, then drains measured usage as browser, AI, artifact, and storage work is observed.
matrixqa / run-001
$ matrixqa inspect --target authorized-url
[01] browser session initialized
[02] journey instructions loaded
[03] evidence capture active
✓ For teams that want predictable controls without pretending every website costs the same to test.
report → screenshots · console · network · timestamps
Live balance movement
Idempotent usage ticks can consume held reserve and then available balance as billable work is recorded.
One clear selling value
The current launch pricing snapshot uses $0.05 per MU. Internal infrastructure cost remains private operational data.
Reconciled at the end
Final measured work, unused reserve, corrections, alpha waivers, and incident treatment are reconciled when the run closes.
The workflow
From one URL to evidence you can act on.
- 01
Reserve before execution
Matrix QA places a bounded hold so a run cannot begin without a safe spending boundary.
- 02
Measure the work
AI usage and authoritative runtime, artifact, and object-storage counters contribute to usage as the run progresses.
- 03
Settle the difference
The terminal result charges measured work, releases unused reserve, or applies the documented infrastructure-failure correction.
Why Matrix Units are not a fixed per-run quote
A browser run is not a single predictable API request. A small smoke journey may need only a few decisions, while a large SaaS workflow can require more navigation, evidence, retries, analysis, or artifact storage. A reserve protects the account while the system learns what the run actually required.
Read how Quick Smoke stays bounded and how Standard Adaptive expands coverage deliberately before comparing run modes.
What happens when a run fails
A failed run is not automatically free and it is not automatically charged in full. Target-access, provider, timeout, and customer-cancelled outcomes preserve authoritative measured work. Matrix QA infrastructure failures and no-work outcomes can reverse provisional customer charges and release unused reserve.
Alpha access can waive the customer charge while measurement continues, so the team can evaluate platform cost without hiding the underlying usage signal.
See the current plan allowance
Free includes 150 MU per month. Starter includes 980 MU per month and adds Standard Adaptive plus governed five-worker collaboration. These are monthly plan allowances, separate from any future top-up purchase flow.
Compare the tiers on the Matrix QA pricing page or review how top-up packages are priced and bounded.
Explore the QA library
View all capabilitiesFind the testing guide that matches your next question.
Questions teams ask
Is one MU the same as one browser run?
No. MU measures billable work, while run cost varies with the mode, decisions, AI usage, evidence, and measured runtime or storage work.
Can the final charge be lower than the reserve?
Yes. Unused reserve is released during terminal reconciliation, and a qualifying Matrix QA infrastructure failure can correct provisional customer charges.
Does Matrix QA expose internal provider cost?
No. Customers receive usage and plan information; raw provider credentials and internal cost details remain operational data.
Turn your next deploy into a proof, not a guess.
Start with one authorized URL. Matrix QA walks the journey and returns the evidence your team needs to decide what ships.