Skip to content
Last updated

Rate limits

Rate limits apply to all Open API key usage and are enforced, not guidance.

Request limits

Limits are enforced per account. Account and record operations and SiftMap requests each have an independent counter:

| Request type | Limit | |---|---|---| | Account and record operations | 120 requests/minute and 5,000 requests/hour | | SiftMap search and map filters | 120 requests/minute and 5,000 requests/hour |

Requests in one group do not use the other group's capacity.

Sustained vs. burst rate

Within each request group, the two caps interact: sustaining the full 120 requests per minute for a whole hour would total 7,200 requests, which exceeds the 5,000-per-hour cap. A process that runs continuously should target an average pace of roughly 80-85 requests per minute (5,000 / 60), reserving the 120-per-minute headroom for short bursts.

If you are building an agentic tool or automated process, build in delays between requests so your sustained throughput stays within the applicable limits, rather than firing requests as fast as possible and relying on retries after a 429. These limits are enough to build a fairly complex, multi-step workflow.

When you exceed a limit

When either the per-minute or per-hour limit is exceeded, the API responds with:

HTTP 429
Request was throttled. Expected available in X seconds.

Your client should recognize a 429, wait the number of seconds in the message (or in the Retry-After header, when present), and retry. This is baseline expected client behavior, not an edge case.

import re, time, requests

def get_with_backoff(url, headers, **kwargs):
    while True:
        r = requests.get(url, headers=headers, timeout=30, **kwargs)
        if r.status_code != 429:
            return r
        m = re.search(r"(\d+)\s*seconds", r.text)
        time.sleep(int(m.group(1)) + 1 if m else 30)

DataSift also reviews API usage regularly, independent of the automated 429 response, and expects fair use; see Acceptable use. For bulk jobs (imports, skip tracing, exports), also follow the Bulk operations limits. Limits may be adjusted as usage patterns evolve; changes are announced in the changelog.

Need more?

If 120 requests/minute or 5,000 requests/hour isn't enough for your process, contact us - we're open to reconsidering these limits based on real use cases.