Skip to main content

Mistral AI

Mistral AI provides a range of open-weight and proprietary models for text and vision, plus audio transcription via Voxtral, and memory embeddings. OpenClaw supports all three use cases with a single MISTRAL_API_KEY.

Requirements

Quickstart

Then verify the default model was set:

Configuration

Environment variable

OpenClaw reads MISTRAL_API_KEY automatically and enables the mistral provider.

Config snippet (LLM provider)

Provider settings

Available models

Use model IDs as mistral/<model-id> (example: mistral/mistral-large-latest).

Audio transcription with Voxtral

Mistral’s Voxtral model handles audio transcription. Configure it under tools.media:
Transcription requests go to /v1/audio/transcriptions.

Memory embeddings

You can use Mistral for memory search embeddings:
The default embedding model is mistral-embed. Requests go to /v1/embeddings.

Troubleshooting

401 Unauthorized
  • Verify your API key is copied in full.
  • Re-run onboarding: openclaw onboard --auth-choice mistral-api-key.
Model not found
  • Confirm the model ID matches exactly (case-sensitive).
  • Check that your Mistral plan includes the requested model.
Rate limits Add multiple API keys for rotation: openclaw models auth add-key --provider mistral.