Skip to main content
Quickstart

From API key to a spoken reply

Create a project key in the dashboard, install the SDK, and stream one turn over the Realtime WebSocket.

1. Get a key

In hub.kupe.in open a project and create an API key. It looks like sk-kupe-... and is scoped to that org and project. Export it (and optionally override the API host):

2. Install

3. Mint a session and connect

GET /v1/me fills in org_id / project_id for API keys, so the SDK constructor can be just Kupe(). Pass voice (name) or voice_id — either one.
The WebSocket speaks OpenAI-compatible JSON. Audio is PCM16 mono at 24 kHz. See Realtime for client and server events. SDK-only walkthroughs: Python and TypeScript.

4. List your agents

Try it in the docs

Every REST endpoint in the API reference tab has an interactive playground. Paste sk-kupe-YOUR_KEY into the Bearer field — requests go to https://x.kupe.in. See API playground.
Coding agents can call the same API over MCP. Install once with http://mcp.kupe.in/mcp — see Kupe MCP.