"Is the new route ready?"
Point a staging key at it, run your real prompts, then move the production key over.
PlaygroundChat · Image · Embeddings
Test the live route before your users do.
Pick an API key and send a message. It runs through that key's published route, on the same models and billing as production, and every reply shows exactly which model answered and what it cost.
What do you want to run through your route?
Under every reply
The receipt, not just the answer.
Each reply carries the facts you need to decide whether a route is ready for production.
- 1Provider and model. Which model in the route actually answered.
- 2Strategy. How the route ranked its models for this request.
- 3Latency. End to end, including any fallback attempts.
- 4Cost. What the request cost: billed by your provider, or taken from credits for Platform Access.
- 5Tokens. Input and output, as the provider counted them.
- 6Raw. The full response, with request ID and route version.
How teams use it
Three questions it answers in a minute.
"Which model is worth it?"
Reorder models or switch strategy, rerun the same prompt, and compare cost and latency side by side.
"Does it handle images and embeddings?"
Generate an image or embed text through the same key, and check sizes, dimensions, and price.
- Same route, models, and billing as production traffic on that key.
- Chat keeps the thread; image and embedding runs are single requests.
- Published versions only: to try a change first, publish it on a route only a staging key uses.
Send your first message.
The guided setup gives you a provider, a route, and a key. The Playground is one click from there.