Trial, limits & errors
Runtime execution burns real compute, so limits are enforced at the start
of every run — from wax signals fire, the Run tab, a schedule, a replay,
or a push. Nothing is throttled mid-run except the wall-clock cap. Every
denial is a structured response with a detail sentence and a link to the
next step; the CLI and the app both show it verbatim.
The runtime trial
Signing up on the Runtime path (or a partner provisioning your workspace for Runtime) stamps a fixed compute grant on the workspace:
| Grant | Value |
|---|---|
| Governed compute | 10 CPU-hours |
| Memory | 20 GB-hours |
| Window | 30 days from signup |
| Concurrent runs | 2 (4 after adding a card in Settings → Billing) |
| Wall-clock per run | 15 minutes |
| Execution tier | shared runtime only |
Whichever of compute, memory, or window runs out first ends the trial. The meter is on Settings → Billing and on the getting-started page; you get a warning at 80%. Everything you built — agents, runs, traces, policies — stays exactly where it is when the trial ends; only new runs are gated.
Custom-image builds during the trial. An agent that declares
dependencies: or execution: isolated is accepted by wax push, but its
image build is parked and it runs on the shared runtime until the workspace
is on a paid plan. The agent's Overview shows "Custom-image builds are part
of paid plans" and wax agents builds <name> says the same.
Limits by plan
| Free (post-trial) | Team | Team + Runtime add-on | Business | Enterprise | |
|---|---|---|---|---|---|
| Runtime execution | — | — | 100 CPU-hr / mo | 100 CPU-hr / mo included | 1,000 CPU-hr / mo included |
| Over the included compute | — | — | hard stop | hard stop until metered billing | per contract |
| Execution tiers | — | — | shared | shared, warm, isolated | all + reserved |
The Runtime add-on is $99 / month on Team and is purchased from Settings → Billing. Business and Enterprise include runtime.
Errors you may see
All of these come back as a JSON body; the CLI prints detail and the link.
error | Status | Meaning | Next step |
|---|---|---|---|
trial_budget_exhausted | 403 | The trial compute or memory grant is used up. | upgrade_url → Settings → Billing |
trial_concurrency_limit | 429 | You already have the maximum number of runs in flight. | Wait for one to finish, or add a card / upgrade |
feature_not_available | 403 | Runtime execution is not on your plan. addon_required: true means the Team add-on unlocks it. | upgrade_url |
compute_quota_exhausted | 403 | This billing period's included compute is used. | upgrade_url |
Run stopped: wall-clock timeout of 900s reached | (run status timeout) | The trial's 15-minute per-run cap. Paid plans set their own via timeout_seconds in waxell.yaml (up to 3600 s) or the Runtime settings panel. | Upgrade, or shorten the task |
trial_tier (blocked_reason on builds) | — | Custom-image build parked during the trial. | Upgrade to build it |
no_active_build | — | Build not finished or failed; runs use the shared runtime meanwhile. | Check the Overview panel: view log, Rebuild |
Where runs execute during the trial
Trial runs always execute on the shared runtime, even for agents that declare an isolated tier. The agent's Overview still shows the declared tier; the run's timeline shows where it actually ran.