Skip to main content
Kilo Code is a VS Code extension for AI-assisted coding. Configure it with a Custom Provider pointing to the AI Gateway to access every model in the catalog, including Anthropic Claude models.

​
Prerequisites

  • Kilo Code extension installed in VS Code (kilocode.kilo-code)
  • Active Orq.ai account
  • Orq.ai API key

​
Setup

Open the Kilo Code panel → Settings → Providers → Custom Provider. Fill the fields:
FieldValue
Providerorq (any string)
Display Nameorqai (any string; appears as prefix in the model selector)
Base URLhttps://api.orq.ai/v3/router
API Key<ORQ_API_KEY>
ModelDeselect all, then activate the models to use (e.g. anthropic/claude-sonnet-4-6, openai/gpt-4o)
Click Submit. In the model selector at the bottom of the Kilo Code panel, pick a model (e.g. orqai / openai/gpt-4o).
Always use the provider/model_id format for model IDs (e.g. anthropic/claude-sonnet-4-6, not claude-sonnet-4-6). Bare IDs return a 400 invalid model format error.

​
Max output tokens

Kilo Code defaults to 32,000 max output tokens. The Custom Provider GUI does not expose per-model token limits. For models with a smaller native cap (e.g. openai/gpt-4o at 16k), open ~/.config/kilo/kilo.jsonc via Kilo Settings → Global Config and add a limit object for the model under the root models key: 
{
  // Merge into your existing kilo.jsonc; do not replace the whole file
  "models": {
    "openai/gpt-4o": {
      "name": "openai/gpt-4o",
      "limit": { "output": 16000, "context": 128000 }
    }
  }
}
Without this, Kilo Code sends a 32k token request. If the upstream model’s native cap is lower, the request will fail with a token limit error.

​
Troubleshooting

The model ID is missing the provider prefix. Change claude-sonnet-4-6 to anthropic/claude-sonnet-4-6, gpt-4o to openai/gpt-4o, and so on.
Add a limit object for the model in ~/.config/kilo/kilo.jsonc as shown in the Max output tokens section above. The GUI does not expose this field.
Verify the Base URL is set to https://api.orq.ai/v3/router. Requests sent to a provider’s own endpoint bypass the AI Gateway and produce no Traces.