Cost controls
Proof starts with live work and AI turned off. Enable them separately, set provider limits, then test one Check first.
Safe defaults
Local setup, linting, type checks, unit tests, builds, Workflow checks, CLI builds, and the documentation site do not create Vercel resources or call an AI provider.
PROOF_LIVE_EXECUTION_ENABLED=false
PROOF_AI_EXECUTION_ENABLED=false
PROOF_MODEL_TIER=budget
PROOF_ALLOW_PREMIUM_MODELS=falseIndependent interlocks
| Variable | Purpose | Default |
|---|---|---|
PROOF_LIVE_EXECUTION_ENABLED | Allows live Workflows and isolated workspaces. | false |
PROOF_AI_EXECUTION_ENABLED | Allows any AI SDK or Gateway request. | false |
PROOF_MODEL_TIER | Selects budget or premium model policy. | budget |
PROOF_ALLOW_PREMIUM_MODELS | Second opt-in required for premium routing. | false |
PROOF_ALLOW_LIVE_VERIFICATION | Allows billable integration verification scripts. | false |
PROOF_ALLOW_BILLABLE_SETUP | Allows creation of a reusable browser snapshot. | false |
PROOF_ALLOW_UNCACHED_BROWSER_RUNTIME | Allows a fresh Chromium download instead of a snapshot. | false |
PROOF_MONITORING_ENABLED | Allows Cron to run saved production tests. | false |
PROOF_MONITOR_AI_ESCALATION_ENABLED | Allows AI to recheck repeated monitor failures. | false |
PROOF_MODEL_MAX_RETRIES | Caps AI SDK retries for each logical model call; run telemetry records the resulting attempts. | 0 |
PROOF_HARD_SPEND_CAP_CENTS | Monthly spending limit. Values above 1000 are rejected. | 1000 |
PROOF_MONITOR_MIN_CADENCE_MINUTES | Rejects schedules faster than the provider Cron wake-up. | 1440 |
Proof reserves a small amount before each billable action and blocks new work at $10. This is not an invoice meter. It works with provider controls. Vercel Spend Management can pause a Pro team. Hobby teams stop at their included limits.
Run the first test
- Use a capped test database and apply reviewed migrations.
- Set up one Clerk development organization and one fixture repository.
- Set catalog and exploration limits to one.
- Enable live execution while AI stays off. Then repeat one saved fixture test.
- Review the isolated workspace lifetime, stored files, Workflow steps, and database use.
- Enable AI for one budget-tier Check only after you set a provider limit.
Operate with AI paused
Saved tests can still run while AI is disabled. Proof skips discovery and does not build new Checks. It marks failures that need an AI recheck as inconclusive. It does not report them as regressions.
Stop conditions
Turn the settings off at once if a Job loops, a deployment does not become ready, isolated-workspace cleanup fails, usage exceeds the test limit, a provider ignores its limit, or a delivery repeats. Keep the Job record. Check the expected retry cost before you try again.
PROOF_AI_EXECUTION_ENABLED=false to stop model calls. Set PROOF_LIVE_EXECUTION_ENABLED=false to stop new live resource execution.