Puffer

Live orchestration console

checking runtime…

OpenAI-compatible API

Generate a sk-puffer-… key for OpenCode and other BYOK clients.

Base URL (OpenAI endpoint)

Chat completions

Key name

Copy now — this secret is shown only once.

Ask the runtime.

Watch intent, routing, model choice, tokens, and cost unfold for a single prompt.

Prompt

Mode Model

Execute models Web search (auto)

Tokens 0

Cost (USD) 0.000000

Tok / sec —

E2E latency —

Model time —

Mode —

Speed breakdown

Pipeline

Intent & difficulty

Models selected

Tokens & cost by model

Node trace

Verification

Answer

idle

POST /v1/orchestrate/stream