Choose a WebSocket API

Choose the Coda JSON or Protobuf WebSocket API, or find the existing Mist WebSocket references.

Use Coda WebSocket v1 for new Coda integrations. For Mist, select an endpoint from the support table below.

Choose a model and protocol

Model	Endpoint	Reference
Coda	wss://api.rime.ai/coda/ws	Coda WebSocket API
Mist v3	wss://users-ws.rime.ai/ws3	JSON reference
Mist v2	wss://users-ws.rime.ai/ws2	JSON reference

Mist v1 and Mist v2 also support /ws3. For all Mist JSON and raw-audio options, see the Mist endpoint support table.

Authenticate from your server

Your server sends Authorization: Bearer YOUR_API_KEY during the WebSocket upgrade. Create a key on the API Tokens page.

Keep credentials on the server. A browser’s WebSocket object cannot set this header, so browser applications need a backend bridge that connects to Rime. See API authentication.

Choose a Coda client

Both encodings support streaming text, persistent connections, context IDs, and cancellation. Neither supplies word timestamps.

Client	Encoding	Start here
Standard WebSocket client	JSON messages with base64 audio	JSON quickstart
Client with Protobuf bindings	Protobuf messages with an audio bytes field	Protobuf quickstart
LiveKit agent	Binary by default; JSON also supported	LiveKit integration

Binary avoids base64 encoding. Decode its Protobuf envelope before playing audio. Use the Coda WebSocket API reference for parameters, lifecycle, and errors.

Mist endpoint support

Mist uses wss://users-ws.rime.ai with the paths below. Use /ws3 for new Mist integrations. Existing Mist v1 and Mist v2 clients can continue to use /ws2.

Model or feature	/ws3	/ws2	/ws
Message format	JSON	JSON	Raw binary audio
Mist v1	✅	✅	✅
Mist v2	✅	✅	✅
Mist v3	✅	❌	✅
Word-level timestamps	✅	✅	❌
Context IDs	✅	✅	❌
Active time-to-first-byte optimization	✅	❌	❌

The Mist v3 JSON reference documents /ws3. The Mist v2 JSON reference documents /ws2; Mist v2 also supports /ws3 with the same JSON message format. /ws2 does not receive new model support.

For raw audio, see the Mist v3 raw WebSocket reference or Mist v2 raw WebSocket reference. The /ws endpoint returns audio bytes without JSON framing, word timestamps, context IDs, or structured error events.

For text buffering and the flush, clear, and eos operations on the Mist JSON endpoints, see the segmentation guide.