Rate limits & usage

Each key has two limits, set in your agreement:

a rate limit: calls per minute;
a monthly quota: calls per calendar month (some agreements have none).

Both apply per key. If one key carries both the Data API and the Analytics API, calls to either count toward the same rate limit and quota. GET /v1/usage shows your limits and this month’s calls.

What counts as a call

A call is counted once it passes authentication, the rate limit and the quota, whatever it returns and however many records it carries:

Request	Counted
A search page, a feed page, a retrieve, a batch of up to 100 ids, jurisdictions, usage	1 call each
A retrieve answered `304 Not Modified`	1 call
A first request that fails validation (400), or a retrieve that finds nothing (404)	1 call
A search or feed continuation request (sent with `cursor`)	1 call if it returns records; not counted if it returns none or fails
A request refused with 401, 403 or 429	Not counted
`GET /v1/health`, `GET /v1/openapi.json`	Not counted (public)

Use limit=100 and batch retrieve: a page of 100 records costs the same as a page of 1.

Rate limit

The rate limit counts calls in the current UTC calendar minute. Above it, calls get 429 rate-limited with "violated_policies": ["minute"] and a Retry-After header with the seconds until the minute ends. Refused calls are not billed, but they still count toward the current minute. Retrying immediately keeps you over the limit; wait for Retry-After. There is also a limit of two concurrent searches per key. A third concurrent search waits up to a second, then gets 429 rate-limited with "violated_policies": ["concurrency"]. Run searches from a small worker pool rather than in parallel bursts.

Monthly quota

The quota counts calls in the current UTC calendar month. Once it is used up, calls get 429 quota-exceeded with "violated_policies": ["month"] until the next month starts. Contact hello@pinakelai.com to raise it. When many requests arrive at the same moment the quota runs out, a few of them can still be served.

RateLimit headers

Every authenticated response reports your key’s limits and what remains, in the format of the IETF draft RateLimit header fields for HTTP (draft 11):

RateLimit-Policy: "minute";q=600;w=60, "month";q=100000;w=2592000
RateLimit: "minute";r=573;t=41, "month";r=95210;t=1555200

Item	Meaning
`"minute"`, `"month"`	The policy name. `month` is absent when your key has no monthly quota.
`q`	The limit: calls allowed in the window.
`w`	The window length in seconds. For `month` it is the length of the current month.
`r`	Calls remaining in the current window.
`t`	Seconds until the window resets.

The headers follow a draft standard, so their exact syntax is outside the stability guarantee until the draft is final. Use them to pace yourself, and use Retry-After on a 429 to decide when to retry. There are no X-RateLimit-* headers.

Python

import re

def remaining(resp) -> dict[str, int]:
    """{"minute": r, "month": r} from the RateLimit header."""
    return {name: int(r) for name, r in re.findall(r'"(\w+)";r=(\d+)', resp.headers.get("RateLimit", ""))}

Checking usage

curl "https://api.pinakelai.com/v1/usage" \
  -H "Authorization: Bearer $PINAKEL_API_KEY"

{
  "period": "2026-12",
  "calls": { "data": 4180, "analytics": 0 },
  "monthly_quota": 100000,
  "rate_limit_per_minute": 600,
  "key": { "name": "Docketing sync (production)", "mode": "live", "products": ["data"] }
}

calls counts this month’s calls for the key, per product. A usage call is itself counted. The customer portal at portal.pinakelai.com shows the same figures with history.

Edge protection

In front of the API, Pinakel’s edge limits very high request volumes per IP address and per key, and blocks clients that send malformed credentials repeatedly. These limits are far above normal use. Responses from the edge can be a 429 or 403 without a JSON body or rate-limit headers; back off and retry later.

Planning your volume

Job	Calls
Check 5,000 docketed marks once a day with batch retrieve	50 a day, about 1,500 a month
Follow the feed every 15 minutes for registers that change about 2,000 records a day	Up to 96 calls a day (polls that find nothing are free), plus batch retrieves for the records you need
Copy a register of 50,000 records (summaries plus batch retrieve)	About 1,000

Test keys have lower rate limits and are not billed.

Get started

Concepts

Data

Use cases

Working with the API

What counts as a call

Rate limit

Monthly quota

RateLimit headers

Checking usage

Edge protection

Planning your volume

​What counts as a call

​Rate limit

​Monthly quota

​RateLimit headers

​Checking usage

​Edge protection

​Planning your volume

What counts as a call

Rate limit

Monthly quota

RateLimit headers

Checking usage

Edge protection

Planning your volume