COST_BUDGET_EXCEEDED

What triggered it

We track LLM and infra cost per account per window (daily / monthly). When the running total reaches your tier’s cap, we pause new generations, deploys, and SSE streams with COST_BUDGET_EXCEEDED (HTTP 429).

The cap depends on your plan and the model you use; the billing page shows your cap and your reset date.

Why we cap spend

The single most-painful failure mode for any AI product is “user got billed $500 because the agent looped.” We hard-cap spend so this can’t happen — even at the cost of telling you “no, slow down” sometimes.

What to try

1. Check your current spend. Billing page → Usage. You’ll see this month’s running total and your cap.

2. Wait for the reset. Monthly caps reset on your billing anniversary. Daily caps reset every 24 hours UTC.

3. Buy a top-up. From billing, click Buy retry for $5 — adds 1 generation outside the cap. Stack as many as you need.

4. Ask a real person. If ordinary use keeps hitting the cap, tell support what you were changing.

Open billing

What if I see this but I’m certain I haven’t generated much?

Two possibilities:

• A loop ran in the background — open the Activity tab in billing to see exactly what happened.
• A teammate’s session generated under your account — happens occasionally with shared credentials. Sign out everywhere and back in.

If you genuinely don’t see the activity that explains the spend:

• Email support@vibecord.dev with your account email
• We’ll pull the per-request log and explain every line

When to contact support

If your spend chart shows numbers you don’t recognize, or if you think your tier’s cap is wrong:

• Discord — post in #help
• Email support@vibecord.dev

We can audit every cost record against the actual LLM / infra logs. If we billed you wrong, we refund — full stop.