TresorTresor API Docs
Workspace Docs

Use Tresor with GitHub Copilot in VS Code

Add Tresor as a Custom Endpoint model in VS Code's GitHub Copilot chat experience.

GitHub Copilot in VS Code can use Tresor through VS Code's Custom Endpoint provider. This setup follows the official VS Code language model flow: open Manage Language Models, add a Custom Endpoint, then save the generated chatLanguageModels.json config.

You need VS Code 1.122 or newer for this flow. VS Code added the Custom Endpoint provider to Stable in version 1.122. If you do not see Custom Endpoint under Add Models, update VS Code first.
This routes the chat model through Tresor for Copilot Chat, tools, and MCP-backed chat flows. It does not replace GitHub Copilot inline suggestions or next edit suggestions. VS Code's BYOK docs note that those still depend on GitHub sign-in and Copilot-hosted model flows.

1. Open the Language Models editor

Open Copilot Chat in VS Code, click the current model picker at the bottom of the chat view, and select Manage Models.

You can also open it from the Command Palette with Chat: Manage Language Models.

2. Add a Custom Endpoint provider

In the Language Models editor, click Add Models and select Custom Endpoint.

VS Code then prompts for the initial provider details:

  • Group name: TresorAI
  • Display name: GLM-5.2
  • API key: your Tresor API key
  • API type: Chat Completions

After that, VS Code opens chatLanguageModels.json so you can finish the model definition.

3. Save the Tresor model config

Use this configuration as the working model definition:

chatLanguageModels.json
[
  {
    "name": "TresorAI",
    "vendor": "customendpoint",
    "apiKey": "${input:chat.lm.secret.<generated-id>}",
    "apiType": "chat-completions",
    "models": [
      {
        "id": "global/redpill/glm-5.2",
        "name": "GLM-5.2",
        "url": "https://api.tresor.co/v1/chat/completions",
        "toolCalling": true,
        "vision": false,
        "maxInputTokens": 128000,
        "maxOutputTokens": 16000
      }
    ]
  }
]

Keep the apiKey reference that VS Code generates for you. The exact ${input:chat.lm.secret.<generated-id>} value is machine-specific, so your suffix will differ from other examples.

The important fields are:

  • vendor: "customendpoint" so VS Code treats it as a Custom Endpoint provider
  • apiType: "chat-completions" because Tresor exposes an OpenAI-compatible chat completions endpoint here
  • id: "global/redpill/glm-5.2" because that is the public model identifier exposed by Tresor's /v1/models API
  • toolCalling: true so the model is available for agent and tool-enabled chat flows
  • vision: false because this model should not be advertised to VS Code as vision-capable
  • url: "https://api.tresor.co/v1/chat/completions" so requests go to Tresor

4. Pick the model in Copilot Chat

Save chatLanguageModels.json, return to Copilot Chat, and select GLM-5.2 from the model picker.

If the model does not appear immediately, restart VS Code. The official VS Code docs call this out as the fallback for newly added Custom Endpoint models.

Once selected, Copilot Chat sends that chat model's requests through Tresor.

See also

Use Tresor with OpenWork