Skip to main content
Run the Kilo Code CLI against GMI Cloud, 500+ models in your terminal, pay-as-you-go pricing. Kilo Code is an open-source AI coding agent for your terminal. GMI Cloud ships as a built-in provider, so connecting takes one keyboard shortcut and an API key: no config files, no subscription, no rate-limit walls on long sessions.

​
Prerequisites

  • A GMI Cloud account at console.gmicloud.ai
  • Node.js 18 or later
  • A terminal running zsh or bash
  • About 3 minutes

​
Step 1. Get your GMI API key

  1. Sign in to console.gmicloud.ai.
  2. Open API Keys at the top right corner.
  3. Create API key and copy it immediately.

​
Step 2. Install Kilo Code

Confirm the install:

​
Step 3. Launch Kilo in your project

​
Step 4. Connect GMI Cloud

  1. Type /models to open the model picker.
  2. Press Ctrl + A to connect a provider.
  3. Scroll to GMI Cloud and press Enter.
  4. Paste your GMI API key and confirm.
  5. Pick your variant.
Kilo stores the key in its local auth store. You only do this once.

​
Step 5. Start coding

Ask for a change directly in the session:
Or run a one-shot prompt from your shell:
Kilo plans the change, edits files, and runs commands, all routed through GMI Cloud.

​
Tips

  • Switch models any time with /models. Kilo is model-agnostic, so match the model to the task: fast variants for quick edits, larger reasoning models for architecture and debugging.
  • Agentic coding leans on tool calling. Pick catalog models with function-calling support for the best results.
  • Set a default model in ~/.config/kilo/kilo.json with "model": "provider_id/model_id", using the exact string shown in the picker.
  • Prefer manual configuration? Add GMI Cloud as an OpenAI-compatible provider with base URL https://api.gmi-serving.com/v1 and any model ID from the LLM catalog. Keep the trailing /v1.
  • Autonomous mode runs without interaction, useful in CI: kilo --auto "run the test suite and fix failures" --timeout 300.

​
Next steps