Skip to main content

Route changes

The older routes keep their original bodies and status names. New starts return a job_id and finish as succeeded, failed or canceled. Store IDs exactly as returned. Deprecated aliases keep their response bodies and HTTP behavior. Their Deprecation and Link headers identify the replacement route. People Intelligence, Company Intelligence, Bulk Intelligence and the Search guides retain the older parameters, tiers, examples and options. Age Verification and the QA Agent have no replacement route. For legacy error bodies, see Handling Errors.

Job status and callbacks

These starts return a task_id. Use it to check status and match callbacks.

Job status

Completed job

Statuses

running is the only status in progress; every other status is final. A failed single-record job is not charged, except for billed tool calls an Atlas run completed successfully before it failed. An Atlas run (xhigh, or scout with graph) that the agents decline still reports completed: its notes start with Declined:, structured_data repeats the input, and nothing is charged. Check notes before you store structured_data. See Declined runs.

Results by job type

Find Email reports find_email, Find Phone reports find_phone, and Reverse Email and Reverse Phone both report reverse_enrichment.

Contact lookups in job status and webhooks

A single-lead contact job returns the lead itself in job status. Its webhook callback wraps the lead in result.leads, with the totals:

Polling

  • Poll each job every 10 seconds or slower; bulk progress updates at most every 10 seconds.
  • Job status shares one organization budget with workflow live status; see Status polling.
  • A 429 or 503 on a status read does not affect the job: wait Retry-After seconds and poll again.
  • To follow many jobs, poll them from one loop that stays within the budget, or pass webhook_url and poll only the jobs whose callback is late.
This example waits for one job and honors Retry-After:

Webhooks

Pass webhook_url to receive the result by POST instead of polling. Outgoing Webhooks documents the payload, signing and retries. Callbacks relate to job status as follows:
  • Job status can read running briefly after the callback arrives.
  • People Intelligence, Company Intelligence and QA Agent callbacks carry no charge; read charge_credits from job status once it reads completed. Their sync endpoints also send the callback when the request has a webhook_url.
  • Contact lookup callbacks carry the final result.cost_credits. Their result differs from job status; see Contact lookups in job status and webhooks.
  • Bulk Intelligence sends a workflow run payload whose run_id equals the task_id. See Bulk Intelligence.
  • Only completed jobs send a callback. If none arrives by your deadline, poll job status.
  • A job can send more than one callback, each with a new event_id; deduplicate on task_id.

Sync calls and retries

A sync endpoint runs the same job and waits for it. If your client times out or disconnects, the job keeps running and is charged when it finishes, and a retry starts and charges a second job.
  • Use the async endpoint for long research: when Sixtyfour is at capacity, a sync call returns 503 and an async job waits.
  • Send dispatch_id (8 to 128 characters) to POST /people-intelligence-async or POST /qa-agent-async to make a start safe to retry. A repeat with the same key, whatever the body, returns the original task_id with status: "RUNNING" and charges nothing while the job is retained; poll it for the real status. Keys are unique per organization across People Intelligence, QA Agent and the Atlas API.
  • Company Intelligence and the contact lookups take no dispatch_id, so a repeated start runs and charges again. A dispatch_id sent to POST /company-intelligence-async comes back in unrecognized_fields.
  • Bulk Intelligence takes an Idempotency-Key header. See Bulk Intelligence.

Retention

Job status is not long-term storage: save each result when its job completes.

Search billing

Legacy deep searches with output_mode: "query_only" create no result file. They charge when results are read with POST /search/query. Legacy lookalike starts are free; reading their results with POST /search/query or exporting them with POST /search/export charges for the rows returned. Viewing lookalike results in the web app is free. Search Jobs instead save and charge for their retrieved results before those results are read. Saved-result exports add no charge. See Credit Costs.