DevelopersIntegrationsKilo Code

Kilo Code with ARK

Add ARK as an OpenAI-compatible provider and run Kilo's agents on open models with tool calling and a 256K-token context.

Kilo Code: An open-source coding agent for VS Code, JetBrains IDEs and the command line.

Before you start

You need an ARK API key: sign in to ARK Cloud and copy it from the dashboard, or use a key from your own ARK deployment and its base URL. The setup below gives each of Kilo's model roles an ARK Cloud model: arklabsai/Qwen3.5-122B-A10B as the main model, arklabsai/Qwen3.6-35B-A3B for subagents and arklabsai/Qwen3.5-4B for small tasks. All three call tools; the 122B and the 35B read up to 256K tokens, the 4B up to 128K.

1. Install

Install
code --install-extension kilocode.Kilo-Code

2. Connect it to ARK

Kilo keeps providers in its config. Add ARK to your global config in ~/.config/kilo, with three models, tool calling and ARK's token limits, and give each role its model:

Kilo config (global)
{
  "model": "openai-compatible/qwen3.5-122b",
  "subagent_model": "openai-compatible/qwen3.6-35b",
  "small_model": "openai-compatible/qwen3.5-4b",
  "provider": {
    "openai-compatible": {
      "options": {
        "baseURL": "https://api.ark-labs.cloud/api/v1",
        "apiKey": "{env:ARK_API_KEY}"
      },
      "models": {
        "qwen3.5-122b": {
          "id": "arklabsai/Qwen3.5-122B-A10B",
          "name": "Qwen3.5 122B A10B (ARK)",
          "tool_call": true,
          "limit": { "context": 262144, "output": 8192 }
        },
        "qwen3.6-35b": {
          "id": "arklabsai/Qwen3.6-35B-A3B",
          "name": "Qwen3.6 35B A3B (ARK)",
          "tool_call": true,
          "limit": { "context": 262144, "output": 8192 }
        },
        "qwen3.5-4b": {
          "id": "arklabsai/Qwen3.5-4B",
          "name": "Qwen3.5 4B (ARK)",
          "tool_call": true,
          "limit": { "context": 131072, "output": 8192 }
        }
      }
    }
  }
}

model is the main model, for coding, chat and reasoning; switch it with the model selector or /models. subagent_model runs the subagents Kilo launches with its task tool, and small_model handles session titles, commit messages and prompt enhancement. The same fields are under Settings → Models.

Set ARK_API_KEY in your environment. In VS Code you can also add the provider from Settings (gear icon) → Providers → Custom provider, with Provider API OpenAI Compatible and the base URL above; token limits and tool calling are set in the config.

Good to know

  • The keys under models, such as qwen3.5-122b, are local names that the roles refer to; id is the model id Kilo sends to ARK.
  • Set small_model yourself: signed in to Kilo Gateway, Kilo otherwise sends those small tasks to its own gateway model instead of your main model.
  • Set limit: without it Kilo does not compact long sessions and assumes 32,000 output tokens.
  • Kilo resolves {env:…} only in trusted locations such as your global config, not in a project's kilo.json.
  • Code completion while you type needs a fill-in-the-middle endpoint, which ARK does not serve: use ARK for chat and agent work.
  • Thinking works with tools: when the agent turns thinking on, the Qwen models think before they call a tool (see Reasoning).
  • These tools send the whole conversation on every turn. ARK's prompt cache speeds up the repeated part of each prompt; every prompt token is still billed.

Next

Read about tool calling and limits and capacity on ARK, or browse other integrations.

Configuration keys follow Kilo Code's official documentation as of November 2026. If the tool has changed since, its docs take precedence.