Skip to main content

​
What is counted

Each successful response is billed the bytes of the rows it sent:
  • On /query: the Arrow data of the rows (the column buffers).
  • On /query/json: the JSON of the rows, which is larger for the same rows. See Arrow vs JSON.
Not counted: HTTP headers, the Arrow schema and metadata (next_block, archive_height, …), IPC framing, and the JSON metadata fields. The bytes depend only on the rows and columns you receive: the number of rows matched and the fields you selected. The size of the range you scanned, the number of filter values and the chain don’t change the price.

​
Minimum per request

Every billed request counts at least 10 KB (10,000 bytes). A response with 3 rows of 100 bytes is billed 10 KB; one with 5 MB of rows is billed 5 MB. On pay-as-you-go, 10 KB is $0.000005: a million requests that return almost nothing cost $5. It matters mostly for very frequent polling; see Estimating costs.

​
What’s free

​
When you’re charged

  • Pay-as-you-go: each response’s cost is taken from your prepaid balance when the response finishes. A response that is already running when the balance runs out completes normally; the next request gets 402 balance_exhausted.
  • Subscription: each response’s bytes count toward the current billing period. The first 1 TB (or your plan’s included data) is covered by the monthly price. Beyond it, overage is billed per GB on your next invoice, up to your overage limit. At the limit, requests get 402 overage_limit_reached.

​
Across chains and keys

Usage from every chain and every key of a project adds up to the same balance or allowance. The dashboard shows it per chain and per key.

​
Seeing your usage

  • Dashboard → Overview: data, spend and balance (or this period’s usage) for any time range, by chain and by key.
  • Dashboard → Billing: the pay-as-you-go statement, or the subscription’s usage meter.
  • API: GET /v1/account returns the key’s plan, balance (pay-as-you-go) or this period’s usage and quota (subscription). It’s free and isn’t rate limited.
Usage appears in the dashboard within about a minute.

​
Making the same data cheaper

  1. Use /query (Arrow), not /query/json: the same rows cost 2–2.5× less.
  2. Select only the fields you use. Leave out logsBloom (256 bytes per row), input, data and accessList unless you need them.
  3. Filter as narrowly as you can on the indexed fields, so fewer rows match.
  4. Don’t fetch a blocks stream for timestamps: select blockTimestamp on logs.
  5. Poll at the block time, not faster, when following the head.
  6. Store what you read. Re-reading finalized history costs the same every time; it never changes.
See Estimating costs for worked numbers.