Feature Request: Add native support for Mistral Code Assistant

[j-jacobi]

Hello Zed team and contributors,

First, thank you for building such a wonderful and performant editor.

The Problem

I'm using "Mistral Code", a code assistant from Mistral AI. The official integrations for this tool are currently limited to VS Code and JetBrains IDEs. This presents a friction point for developers who prefer using Zed but need to work within the company's chosen toolset.

https://mistral.ai/products/mistral-code

Proposed Solution

I propose adding Mistral Code as a "first-class citizen" in Zed's list of AI providers. Presenting the user with a simple "Mistral Code" option in the provider list that only requires them to enter their API key.

Feasibility and Current Workaround

Mistral Code exposes an OpenAI-compatible API. I have successfully configured it using Zed's generic "OpenAI-Compatible" provider.

For anyone else looking to achieve this manually, here are the steps that worked for me:

• Open the Agent Panel in Zed

• Click on the currently displayed model name, then click "Configure".

• Click on + Add Provider.

• Select the OpenAI-Compatible provider type.

• Set the provider name to Mistral Code. (Or any value you want)

• For the API URL, enter: https://codestral.mistral.ai/v1

• For the API Key, paste the token found in your Mistral Code configuration file after installing the Mistral Enterprise Code extension on VSCode (typically located at ~/.mistralcode/config.json).

• Click Add Provider.

• Edit your Zed settings.jsonfile to add the list of available models to the provider you just created.

• Here is my language_models block that works for Mistral Code needs to be added to settings.json to make the models available in the Agent Panel.

"language_models": {
  "openai_compatible": {
    "Mistral Code": {
      "api_url": "https://codestral.mistral.ai/v1",
      "available_models": [
        {
          "name": "mistral-medium-2508",
          "display_name": "Mistral General Model",
          "max_tokens": 128000,
          "max_output_tokens": null,
          "max_completion_tokens": null,
          "capabilities": {
            "tools": true,
            "images": true,
            "parallel_tool_calls": true,
            "prompt_cache_key": false
          }
        },
        {
          "name": "devstral-medium-2507",
          "display_name": "Devstral Coding Model",
          "max_tokens": 128000,
          "max_output_tokens": null,
          "max_completion_tokens": null,
          "capabilities": {
            "tools": true,
            "images": false,
            "parallel_tool_calls": true,
            "prompt_cache_key": false
          }
        },
        {
          "name": "codestral-2508",
          "display_name": "Codestral Code Completion",
          "max_tokens": 256000,
          "max_output_tokens": null,
          "max_completion_tokens": null,
          "capabilities": {
            "tools": true,
            "images": false,
            "parallel_tool_calls": true,
            "prompt_cache_key": false
          }
        }
      ]
    }
  }
}

Benefits

Increases Zed's adoption

Enhances user choice

Improves user experience

I am opening this issue to start a discussion if the above settings can be interesting to anyone else

Thank you for your consideration!