Plan philosophy

This is a companion to the plan tier limits reference. The reference tells you what each tier includes; this page explains why.

The principle

A pricing model that throttles you at the worst moment — your build’s red, you’re trying to debug, you hit the cap — is a hostile one. We designed the tiers so the caps line up with growth rather than punishment.

The two questions that drove the catalog:

What lets a Free user actually be productive (not just trial)?
Where does each paid tier earn its money without forcing a worse feature on everyone below?

Free is generous on schedules

Free includes 3 scheduled runs. That’s enough to set up a daily smoke check, a weekly regression, and a one-off CI gate without ever paying. Most “free trial” patterns put schedules behind the first paywall — we don’t, because tests without scheduling are barely tests.

The cap that hits Free users first is usually monthly cloud-runs (200), not the schedule count.

BYOK unlocks unlimited AI on every tier

Every plan — Free included — lets you connect your own OpenAI / Anthropic / Groq / Gemini key. When you do, AI creations and visual compares run on your account, not against Marriska’s quota.

A Free user with BYOK behaves like an unlimited-AI tier for the AI side. The cloud-runs quota still applies (because that’s compute we pay for), but the AI side is decoupled.

Why: AI costs scale with use. If we cap AI per tier, heavy users either overpay (we charge enough to cover worst-case) or hit a wall. BYOK breaks the bind — you pay the LLM provider directly at provider rates, we don’t lose money, you don’t get throttled. See the BYOK concept for the full story.

Parallel executions scale with tier

Free / Starter get 1 parallel execution per org. Pro 2, Team 5, Enterprise unlimited.

Parallelism is the most expensive cap to raise on our side — each slot is a Playwright instance burning CPU. It’s also where heavy users feel real productivity gain (5 tests running concurrently in CI is 5× faster, full stop). Paying tiers get more slots; Free is designed for one-at-a-time experimentation.

Retention scales with seriousness

Run history retention: Free 14 days, Starter 90 days, Pro 1 year, Team / Enterprise unlimited.

Free is “I’m trying this out” — last two weeks of runs is enough to notice trends. Starter is “I’m using this for a side project” — 90 days covers a release cadence. Pro is “this is my real CI” — 12 months covers an audit window. Team / Enterprise compliance teams genuinely need indefinite retention; we don’t charge for storage proportional to what they store.

Seats and team features pay for themselves

Multi-user collaboration (shared projects, org roles, audit logs, SSO) is gated at Team. Up through Pro, every project is owned by one user — fine for individuals and small teams.

Team is where multi-user workflows actually open up:

Shared projects — every member sees the project, not just the creator
Org roles — Owner / Admin / Member / Viewer with real permission boundaries
Google Workspace SSO — one click to add a new hire

Enterprise adds SAML SSO, audit logs, and SLA — features procurement asks about, not features individual users need.

The pricing-page promise

The reference page lists the numbers; this page promises the spirit behind them:

You can be productive on Free. It’s not a 14-day trial in disguise.
You pay for compute and parallelism, not for AI.
Caps line up with growth, not with arbitrary paywalls.
No quota cliffs. Visible warnings before you hit a limit; no silent failures.