Skip to main content

Kilo Code

Kilo Code is an open-source AI coding extension for VS Code (or any other VS Code-based editor like Cursor, CodeSandbox, etc.). Because Nebul's Inference API is OpenAI-compatible, you can connect Kilo Code directly to your Nebul project without routing code through third-party model providers.

Prerequisites​

Installation​

Install the Kilo Code extension from the VS Code Marketplace:

  1. Open VS Code.
  2. Go to Extensions (Ctrl+Shift+X / Cmd+Shift+X).
  3. Search for Kilo Code.
  4. Click the dropdown arrow next to Install and select Install Pre-Release Version.

The pre-release channel is the current recommended distribution for Kilo Code. For Open VSX, VSIX installs, and troubleshooting, see the Kilo Code installation guide.

Configure Nebul​

Kilo Code supports OpenAI-compatible endpoints through the OpenAI Compatible provider. Configure it in the extension settings:

  1. Open Kilo Code in VS Code.
  2. Click the gear icon in the extension UI to open Settings.
  3. Go to the Providers tab.
  4. Select OpenAI Compatible as the API provider.
  5. Set the following fields:
SettingValue
Base URLhttps://api.inference.nebul.io/v1
API KeyYour Nebul API key from AI Studio
ModelA model ID from the Model Catalog, for example zai-org/GLM-5.1-FP8
  1. Under Model Configuration, set Context Window and Max Output Tokens to match your chosen model's capabilities. Kilo Code uses these values for context management and compaction.
  2. Save your settings.

For models that support tool calling — required for file edits, terminal use, and agent workflows — enable tool support in the model configuration if the option is available.

Alternative: custom provider in config​

If you prefer file-based configuration, add Nebul under the openai-compatible provider in kilo.jsonc:

json
12345678910111213141516171819202122
{
  "$schema": "https://app.kilo.ai/config.json",
  "model": "openai-compatible/zai-org/GLM-5.1-FP8",
  "provider": {
    "openai-compatible": {
      "options": {
        "apiKey": "{env:NEBUL_API_KEY}",
        "baseURL": "https://api.inference.nebul.io/v1"
      },
      "models": {
        "zai-org/GLM-5.1-FP8": {
          "name": "GLM 5.1 FP8",
          "tool_call": true,
          "limit": {
            "context": 128000,
            "output": 16384
          }
        }
      }
    }
  }
}

Set NEBUL_API_KEY in your environment, or replace {env:NEBUL_API_KEY} with your key directly. See Custom models for the full config reference.

Start coding​

Open a project in VS Code, pick an interaction mode (for example Code or Architect), and send your first task. Requests are sent to Nebul using the model and API key you configured. Usage appears under the API key's project in AI Studio.

Troubleshooting​

  • Invalid API key — Confirm the key is from the correct AI Studio project and has not been revoked.
  • Model not found — Use the exact model ID from the Model Catalog. Model availability depends on your project's enabled models.
  • Connection errors — Use the base URL https://api.inference.nebul.io/v1, not the OpenAI default endpoint.
  • Tools or edits not working — Choose a model that supports tool calling and enable it in model configuration.

For provider-specific setup details, see Using OpenAI-compatible providers with Kilo Code.

info

Do not use the OpenAI Compatible provider for Azure OpenAI GPT-5 deployments — Azure requires Kilo Code's native azure provider. Nebul uses the standard OpenAI-compatible chat completions schema.

Last updated on