Pricing
Credit-based pricing, free 1,000 credits/month.
Brime uses a single credit balance per account. Each call costs a fixed number of credits, set by the endpoint.
Credit table
| Endpoint | Credits | Notes |
|---|---|---|
/v1/search | 0.5 | Single tuned tier: SERP snippets + short synthesised answer; depth is ignored |
/v1/extract | 1 per URL | Charged only for successfully extracted URLs; failed URLs are free |
/v1/research | 5 | Flat per call, sync with optional SSE (~30-90 s). Tavily-compatible deep research costs 15 |
Errors do not charge credits. Cache hits charge the original endpoint's cost.
Plan tiers
Free
1,000 credits / month
No credit card. Perfect for prototyping and small side projects.
Pay as you go
$0.005 / credit
No minimum and no monthly commitment.
Pro
Monthly plans
5,000 credits for $29, 15,000 for $49, 40,000 for $99, or 100,000 for $199 per month.
Current prices are always listed at brime.dev/pricing. Usage beyond a Pro plan's monthly credits is billed per credit; see that page for the current rate.
How credits are deducted
Credits are debited after a successful response, not before. The flow:
- Worker handler accepts and validates your request
- Endpoint executes (search, extract, or research)
- On
2xx, Brime emits an internal usage logs event withcredits_used - A cron worker rolls usage events into your account balance every minute
This means a single high-cost call won't suddenly block other in-flight requests — billing is asynchronous. We do enforce a soft-cap when the balance drops below zero (next request returns insufficient_credits).
Tracking usage
Every response includes:
{
"credits_used": 1.5,
"request_id": "req_..."
}Plus headers:
X-Credits-Used: 1.5
X-RateLimit-Remaining: 9981
X-RateLimit-Reset: 1715000000Use the dashboard for historical analytics: usage by endpoint, by key prefix, by hour/day.
Roll-over
Credits do not roll over between billing periods.
Refunds
If you observe an erroneously charged request (e.g., a 5xx that returned 2xx upstream), file a ticket with the request_id — we'll inspect logs and refund manually.