Skip to main content
POST
JavaScript

Authorizations

Authorization
string
header
required

Bearer token via Authorization header

Headers

Idempotency-Key
string | null
Required string length: 1 - 128

Body

application/json
deviceId
string<uuid>
required

The ID of the device to run the task on.

task
string
required
Minimum string length: 1
accessibility
boolean
default:true
apps
string[]
continueOnFailure
boolean
default:false
credentials
PackageCredentials · object[]
displayId
integer
default:0

The display ID of the device to run the task on.

executionTimeout
integer
default:1000

Maximum agent execution time in seconds (1–2700).

Required range: 1 <= x <= 2700
files
string[]
humanize
boolean
default:false

Use variable timing and gesture paths instead of instant events where supported by the device.

llmModel
string
default:openai/gpt-6-luna

The LLM model identifier to use for the task (e.g. 'openai/gpt-6-luna')

maxSteps
integer
default:100
outputSchema
Outputschema · object | null
reasoning
boolean
default:true
recordingEnabled
boolean
default:false

Record device video for the whole task and persist a retrievable reference

subagentModel
string
default:openai/gpt-6-luna

LLM model used by sub-agent roles: executor, app_opener, structured_output

systemPrompt
string | null

Optional custom behavioral overlay applied on top of the agent's default system prompts. Never echoed back in responses or errors.

Maximum string length: 32000
temperature
number
default:0.5
deprecated

Deprecated and ignored. Sampling behavior is controlled by the model provider.

vision
boolean
default:false
vpnCountry
enum<string> | null
Available options:
US,
BR,
FR,
DE,
IN,
JP,
KR,
ZA

Response

Successful Response

id
string<uuid>
required

The ID of the task

status
enum<string>
required

The status of the task (queued or created)

Available options:
prepared,
queued,
created,
running,
cancelling,
completed,
failed,
cancelled
streamUrl
string | null

The URL of the stream (null when queued)