openclaw - 💡(How to fix) Fix [Bug]: Mistral direct API works, but OpenClaw chat runs fail with 400 status code (no body) and appear to hang [2 comments, 2 participants]

{ "subsystem": "agent/embedded", "event": "embedded_run_agent_end", "isError": true, "error": "400 status code (no body)", "failoverReason": null, "model": "mistral-large-latest", "provider": "mistral", "rawErrorPreview": "400 status code (no body)", }

Steps to reproduce

Steps to reproduce

1. Configure Mistral as a provider in OpenClaw with a valid API key.
2. Set the active/default model to mistral/mistral-large-latest.
3. Restart the OpenClaw gateway so the updated provider config is loaded.
4. Send a normal user message in chat (for example in Telegram / direct chat).
5. Observe that the chat appears to hang or does not return a proper reply.
6. Check the OpenClaw logs.
7. Observe entries like:
   • provider: "mistral"
   • model: "mistral-large-latest"
   • error: "400 status code (no body)"

Control test

1. Use the same Mistral API key directly against: https://api.mistral.ai/v1/chat/completions
2. Send a very simple prompt such as: Reply with exactly: MISTRAL_OK
3. Observe that the direct API call succeeds.

This shows that the credential and model are valid, while the OpenClaw runtime path still fails.

Expected behavior

If the same Mistral model and API key work directly, OpenClaw should also be able to complete normal chat runs successfully.

If the provider rejects something in the request, OpenClaw should ideally:

surface a more specific error and not leave the user with what looks like a hanging chat

Actual behavior

API key is configured and valid direct call to Mistral works provider/model path in config is structurally correct OpenClaw runtime path fails with 400 status code (no body) This suggests a request-shape / adapter / runtime compatibility problem in the OpenClaw -> Mistral path.

OpenClaw version

v2026.3.28

Operating system

Linux

Install method

No response

Model

mistral-large-latest

Provider / routing chain

openclaw

Additional provider/model setup details

No response

Logs, screenshots, and evidence

Fix Plan

The fix involves modifying the OpenClaw request to match the expected format of the Mistral API.

• Update the OpenClaw configuration to include the necessary headers and parameters for the Mistral API.
• Modify the OpenClaw request to use the openai-completions endpoint correctly.

Example code changes:

# Update the request headers to include the API key
headers = {
    'Authorization': f'Bearer {api_key}',
    'Content-Type': 'application/json'
}

# Update the request parameters to match the Mistral API expectations
params = {
    'model': 'mistral-large-latest',
    'prompt': user_message
}

# Send the request to the Mistral API
response = requests.post('https://api.mistral.ai/v1/chat/completions', headers=headers, json=params)

• Verify that the api_key variable is set to the correct value and that the user_message variable contains the user's input.

Verification

To verify that the fix worked, send a test message through the OpenClaw chat interface and check the response. The chat should no longer hang, and a proper reply should be returned.

• Check the OpenClaw logs for any error messages.
• Verify that the Mistral API is returning a successful response (