Skip to main content
POST
People Intelligence (Async)

Headers

x-api-key
string | null

Body

application/json
lead_info
Lead Info · object
required

Single lead to process. Provide any combination of name, email, company, linkedin, etc.

struct
Struct · object | null

Mapping of output field name to natural-language description of what to extract. A field that asks for an image (a headshot, a logo, a screenshot of a page) is returned as a permanent image URL on api.sixtyfour.ai/v1/media, hosted by Sixtyfour; embed it directly.

research_plan
string | null

Optional natural-language plan that guides the research agent.

tier
enum<string> | null

Quality and cost tier for the research agent.

Available options:
low,
micro,
medium,
high,
xhigh,
scout
webhook_url
string<uri> | null

HTTPS URL that receives the result payload when the async job completes.

Required string length: 1 - 2083
field_confidence
boolean | null

If true, return each requested field's confidence (0-100), justification and source URLs, nested like the request.

include_workspace
boolean | null

For Atlas runs (xhigh, or scout with graph), include workspace files alongside structured output (default: true). Set false to omit them. Job status reads them from the investigation, so it shows the investigation's current workspace, including what a later run on it changed. Saved images appear as attachment records; requested image fields carry their permanent /v1/media URL, and the investigation's /v1/atlas file API serves any attachment. Ignored by other tiers.

investigation_id
string | null

For Atlas runs (xhigh, or scout with graph), continue an existing investigation instead of creating one. The run reuses that investigation's graph and workspace, so a follow-up builds on the earlier research rather than starting over. Ignored by other tiers.

investigation_name
string | null

For Atlas runs, the title of a newly created investigation. Ignored when resuming via investigation_id; when unset the title is derived from lead_info.

relentless
boolean
default:false

For Atlas runs with target fields: research until the budget runs out. By default a run may stop once every requested field is answered, charging the turns it used (at least one unit).

max_credits
integer | null

For Atlas runs, how much research to buy, in credits (default: one unit of the tier price, 250 credits on xhigh), up to 10 units. On multi-agent runs every 250 credits buy 800 team turns (every model response by any agent in the run), used in full unless the research stops yielding: 300 buys 960.

max_minutes
integer | null

For Atlas runs, the longest the run may research, in minutes of working time: waits to retry a model call or an attempt do not count. When it is reached the run writes up what it found and completes, charging the credits its turns used, rounded up to the next 50, instead of the whole budget. Writing up can take a few minutes past the limit. Default: no limit.

Required range: 10 <= x <= 1440
dispatch_id
string | null

Async endpoints only: a client-generated idempotency key for the dispatch. Reusing the same key for the same organization returns the existing job instead of starting and billing a duplicate.

Required string length: 8 - 128
exclusive_run
boolean
default:false

Atlas runs on an existing investigation only: refuse with 409 while that investigation already has a run starting or running, instead of starting and billing a second one.

Response

Successful Response

Async people-intelligence kickoff.

Atlas jobs (xHigh or Scout Graph) expose their workspace immediately. The fields stay optional because the same endpoint supports tiers that do not create an investigation.

task_id
string
required

Async job ID. Poll GET /job-status/{task_id} for status and results.

status
string
required

Initial job status.

Allowed value: "RUNNING"
investigation
InvestigationReference · object | null

Atlas runs only: the workspace created for this job.

unrecognized_fields
string[] | null

Request fields we do not recognize and therefore ignored, so an instruction that silently did nothing is visible rather than assumed to have applied. Absent when every field was understood.