Skip to main content
GET
Platform choices for the per-mode subagent seat grid.

Authorizations

x-access-token
string
header
required

JWT token passed in x-access-token header

Response

Seat grid options.

success
boolean
modes
enum<string>[]
Available options:
advanced,
expert,
mythical
defaults
object

all = the platform model per role slot (plannerModel, coderModel, testerModel, writerModel, agenticLoopModel, preflightModel, designerModel, researcherModel, specWriterModel, planReviewerModel, judgeModel, vision — null: no platform default) plus conductorModel and conductorFallbackModels. Each mode carries every role slot plus coderUpgradeModel (the coder for harder tasks) and coderVisualModel (the coder for games and simulations); a level is null when the mode has none or it equals the mode's coder. specWriterModel follows the mode's planner.

judgeVisual
object

Per mode, the judge used when grading a visual (games and simulations) build — display hint only.

platformEfforts
object

Platform reasoning effort per role slot and model, used when the twin sets none and the run has no mode.

platformTierEfforts
object

Platform reasoning effort per mode and role slot, one level per mode (advanced medium, expert high, mythical xhigh), used when the twin sets none. In a run with a mode it wins over platformEfforts. Empty when the per-mode platform defaults are switched off.

efforts
object

Per catalog model value: levels = the efforts that change what is sent for that model (empty: effort has no effect), aliases = every other effort mapped to the offered level it runs as. A model that is absent supports the full list.