An empty thread now opens with three starter cards: send your first real request, check whether a cheaper model holds up, run the demo.
Three starting points
The starter cards gave an empty agent thread a practical first task: send a real request, compare the cost of another model, or explore a demo. Each card provided a starting point for a different task.
Send your first request
Choose a project, create an application key, and send a request to the Stimulir inference gateway. Open Traces in the same project to inspect the response and any recorded latency, tokens and cost. The first-call guide and developer quick start include the setup steps and examples.
Compare before changing models
Use the same task cases and evaluation criteria when comparing models. Inspect quality alongside cost and latency; a cheaper response is useful only if it still meets your requirements. See Evaluation for controlled comparisons and Improve for a trace-based improvement workflow.
Explore without confusing a demo with your data
A walkthrough can explain the interface before you connect an application. Use Watch mode, then verify the workflow with your own project and evidence. A demo result is not a measurement of your production system.
About this entry
This expands the original 16 September 2026 announcement. Its original deployment record was not retained. Card labels and availability can differ in the current Console; the linked guides describe the supported next steps.