Why credits?
Not all data is created equal — our credit model reflects the real cost of retrieving and processing financial data for AI agents.
Credits are fairer than flat per-call billing
A simple metadata lookup (1 credit) costs far less to serve than a full SEC filing retrieval across hundreds of pages. Flat per-call billing would force us to overcharge for simple queries or lose money on complex ones. Credits align cost with value.
How credit consumption works
| Endpoint Category | Credits |
|---|---|
| Lightweight metadata (search_, list_, get_) | 1 |
| Page map / keyword search (read_source_outline, search_keyword_in_source) | 2 |
| Deep read (read_source_pages) | 1 per page retrieved |
| Cross-period search (search_cross_period) | 1 per fiscal period |
See your credit usage in real time
Every API response includes HTTP headers showing your consumption:
X-Credits-Used— credits used by this callX-Credits-Remaining— the account’s total remaining across all three pools, after this call. It is a number on every response, success or error.X-Credits-Split— which pools paid for this call, in the order they are consumed:daily:3,monthly:2
Your agent can read these headers programmatically to track and optimize credit usage. See Rate Limits & Credits for full documentation.
Errors never cost credits
Authentication failures (401), bad parameters (400), rate limits (429), and server errors (500) consume zero credits. Only successful data responses (HTTP 200) are metered. We want you to explore freely.