Give your agent
a next move.
A goal, the current state, and two to eight allowed moves. Trio-Flash helps GUI agents and interactive apps choose a next action your code can inspect.
Planned limited-preview price: $0.042 / million billed input tokens · Output free · $5 minimum top-up
Public access is still gated. The Playground and API keys open only after live service and billing validation.
What should the agent do next?
- 01Open the Export menu
- 02Open report filters
- 03Wait for more information
Illustrative structured UI state, not a screenshot. Change the state to inspect a request; no model has been called here.
{
"model": "trio-flash-preview",
"task": "The user asked to export the current report as CSV. Choose the next safe UI action from the available controls.",
"state": {
"app": "Reports",
"screen": "Report details",
"visible_controls": [
"Export menu",
"Filters",
"Help"
]
},
"choices": [
{
"id": "open_export_menu",
"description": "Open the Export menu"
},
{
"id": "open_filters",
"description": "Open report filters"
},
{
"id": "wait",
"description": "Wait for more information"
}
]
}Run this request from the Playground above when the limited preview opens.
Bring the current state.
An accessible UI tree, a game turn, or a workflow described in text or JSON. You define the task.
Define the next moves.
Supply 2–8 choices with your own IDs and descriptions. Keep the actions meaningful to your application.
Make the decision useful.
Receive a choice and probabilities. Your code decides when to act, ask for review, or wait.
Small decisions.
Room to build.
Start with a task you can check: selecting a GUI agent’s next control, choosing a Falling Blocks move, or routing a request. When the Playground opens, try your own states and choices.
The API returns a choice ID, probabilities, a decision status, and usage. Review uncertain choices before turning them into actions.
This is a request example, not a measured result. The current API accepts text or JSON state; it does not read screenshots. When the Playground opens, the platform will show the current input limit and availability.
Use the single Playground entry above when the limited preview opens.
curl https://platform.machinefi.com/api/flash/v1/decisions \
-H "Authorization: Bearer $TRIO_FLASH_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $(uuidgen)" \
-d '{
"model": "trio-flash-preview",
"task": "The user asked to export the current report as CSV. Choose the next safe UI action from the available controls.",
"state": {
"app": "Reports",
"screen": "Report details",
"visible_controls": [
"Export menu",
"Filters",
"Help"
]
},
"choices": [
{
"id": "open_export_menu",
"description": "Open the Export menu"
},
{
"id": "open_filters",
"description": "Open report filters"
},
{
"id": "wait",
"description": "Wait for more information"
}
]
}'$0.042
per million billed input tokens · output free
A 400-token request costs $0.0000168. A 1,000-token request costs $0.000042. Actual cost depends on your task, state, and choices.
Completed judgments, including review or abstain, are billed. Failed requests and idempotent replays are not. Private template overhead is excluded.
This is a limited-preview rate. Pricing and service limits may change before general availability; the Playground will show current terms before paid use.
A first glimpse of Trio’s
decision intelligence.
Trio-Flash brings the decision side of Trio to developers through a hosted API. Start with a small, testable problem. See what it can help you build.
Preview capabilities vary by task. Test on your own examples before relying on a decision. Video and sensor understanding are not inputs to this API.