Billing & Usage
Shiplight Cloud bills for what you actually use, priced directly in dollars. There is no credit currency to convert — your plan gives you an allowance in dollars, usage draws it down in dollars, and the usage view reports dollars.
Owners manage billing at app.shiplight.ai/settings/billing. Every member can see usage.
What's metered
Three meters, each measured in its own unit:
| Meter | Unit | Covers |
|---|---|---|
| Tokens | Millions of tokens (MTok) | LLM calls through the LLM proxy |
| Compute | Machine-minutes | CI runners and hosted sessions, priced by machine size |
| Storage | GB-months | Persistent workspace disk, beyond the amount included with your seats |
Token usage counts every token the provider reports: input, output, thinking, cache reads, and cache writes. Models are priced at one of three tiers — Lite, Standard, or Pro — listed per model under LLM proxy.
Storage carries no additional charge by default — only workspace disk beyond your included allowance is billable, and test artifacts are not metered at all.
Per-unit rates depend on your plan or contract. The authoritative figures for your organization are the ones in Usage and on the Subscriptions page.
Usage is priced using the rates in effect at the time it occurred, so a later price change never retroactively reprices past usage.
Two balances
Your organization's funds sit in two buckets, and they behave differently:
| Included allowance | Top-up balance | |
|---|---|---|
| Where it comes from | Granted by your plan each billing cycle | Bought separately |
| At renewal | Resets — unused amounts do not roll over | Rolls over, never expires |
| Can go negative | No | Yes |
Usage draws from the included allowance first, then the top-up balance.
When you run out
New work is admitted while the combined balance is above zero and blocked once it isn't. Work that has already started is always allowed to finish, which is why the top-up balance can end up slightly negative — the final cost of an in-flight run settles after admission.
At renewal, only the included allowance resets. A negative top-up balance carries forward until a top-up or auto-recharge clears it.
Your test results and artifacts are unaffected — they aren't metered. Persistent workspace storage becomes read-only, and is released if the balance isn't restored before the grace period ends (90 days by default).
Plans
There are four plans: Free, Pro Monthly, Pro Yearly, and Enterprise. Each grants an included allowance every billing cycle — automatically renewed on Free, discounted against twelve monthly cycles on Pro Yearly, and set by contract on Enterprise. Current prices and allowances are on the Subscriptions page.
Every plan includes the full feature set — AI actions, natural-language steps, auto-healing, failure recovery, MCP access, and skills. Paid plans add auto-recharge, audit logs, and the ability to receive recordings from non-members; Enterprise adds an SLA and dedicated support. Organization members are capped per plan.
Change plans at Subscriptions. While a paid subscription is active the plan is locked — cancel to switch, or schedule the change for the end of the current period.
Adding funds
Top-up adds to the balance that rolls over. Choose $10, $25, $50, or $100, or enter a custom amount; the minimum is $5.
Auto-recharge tops up automatically when your balance falls below a threshold you set, charging your saved payment method. Defaults are a $5 threshold and a $20 recharge. You can cap how much auto-recharge is allowed to spend per day and per month, and the page shows how much of each cap you've used.
Both require an existing paid subscription, since they charge a payment method on file.
Invoices
Invoices lists every charge — subscriptions, top-ups, and auto-recharges — filterable by type and date, each linking to its hosted receipt.
Usage
app.shiplight.ai/usage is visible to every member, not just owners. It defaults to the last 30 days and offers presets of 7, 14, 30, and 90 days.
You get total spend plus a breakdown across tokens, compute, and storage, and a daily chart stacking cost by category. Filter by date range, meter, and member.
Export downloads exactly what's on screen as CSV, with one row per usage event: timestamp, category, provider, pricing tier, machine type, quantity, unit, cost, and what the usage was for.
A separate LLM usage breakdown reports token counts whether or not Shiplight billed them, labelling each row Proxy, BYOK, or Custom endpoint. Calls made with your own provider key appear as counts with no cost, so you can see what your tests consumed even when you're paying the provider directly — see LLM Providers.
This breakdown covers test runs only, not every API call, and attributes usage to the run's start date — so its totals won't line up exactly with the cost chart at the edges of a date range. It is also hidden while a member filter is applied.
What the member filter can and can't attribute
Filtering by member only matches usage driven by someone's personal API token. Usage from organization tokens, CI runs, and hosted sessions is billed to the organization rather than a person, so it disappears when you filter by member. Leave the filter off for a complete picture.
Analyze usage with Shiplight Cloud
The Usage page shows the underlying numbers. /shiplight cloud lets your coding agent query the analytics API, explain what is driving token and compute usage, and turn the findings into specific cost-saving recommendations.
For example:
/shiplight cloud analyze token and compute usage for the last 30 days and suggest ways to reduce costThe analysis covers:
- LLM usage — calls and tokens by operation, provider, model, routing method, and proxy tier. It can highlight expensive AI-heavy steps, model-driven conditions and waits, and low locator-cache hit rates.
- Runner compute — VM minutes and cost, machine size, tests per job, and achieved concurrency for Shiplight-hosted CI runners. It can identify oversized runners and workflows split into too many small jobs.
- Cost efficiency — token and compute cost per test, with the evidence behind each recommendation so you can decide whether changing a model, step, cache, runner size, or job layout is worthwhile.
Token analysis includes usage reported by proxy, BYOK, and custom-endpoint test runs, although only proxy calls are billed by Shiplight. Compute analysis is available for Shiplight-hosted runners. The command is read-only and requires a full-access token or the analytics:read scope; see API Tokens.
app.shiplight.ai/account/usage gives a quick view of the current allowance and what's left. Despite its name it reports organization-level figures.
Enterprise
Enterprise is a contract rather than a self-serve plan, and it bills differently in ways worth knowing:
- Usage beyond the included amount does not block — it accrues as overage and is invoiced separately.
- Invoicing is handled by your account manager. There is no top-up, no auto-recharge, and no credit balance.
- Contract terms, billing contact, invoice terms, and PO number appear on the Billing page.
Related
- Organization — roles and who can manage billing
- API Tokens — LLM proxy usage and attribution
- CI Runners — what drives compute usage