Playground

Two playgrounds live in the dashboard. Both use your account session (not a long-lived API key you paste yourself).

​

Model playground

Open Playground. Pick a Ready or Sleeping model. The UI switches endpoints to match the pipeline, including:

• Chat / completions
• Embeddings
• Images
• Speech and transcriptions
• Text and token classification

Use it to sanity-check a launch, then copy the request into Examples or an integration. Navbar Playground in these docs points at this route.

​

Agent playground

Open Agent playground after you deploy an agent. It streams the same AG-UI events as POST /v1/agent/{agent_id}/run. Tools, MCP, and knowledge bases come from the saved agent, not from extra request tools.

​

Next steps