AI-Assisted Coding

Serverless Inference with Cline

This tutorial walks you through using Kimchi's Serverless Endpoints with Cline, an agentic coding tool available as a standalone CLI and as a VS Code extension. You'll configure Cline to use open-source LLMs via Kimchi, providing a cost-effective setup for AI-assisted development.

Overview

By the end of this tutorial, you'll be able to:

Authenticate Cline against Kimchi as a model provider
Use MiniMax M2.5 for code generation and execution tasks
Run Cline interactively in your terminal or headlessly in scripts and CI/CD pipelines

📘

Serverless Endpoints work with any OpenAI-compatible client. This tutorial covers Cline, but the same Kimchi base URL and API key work with OpenCode and other compatible tools.

Prerequisites

A Kimchi API key — Go to Kimchi → API Keys and click Create API key. Copy the key and store it securely.
Node.js 20 or higher — Check your version with node --version
Cline CLI installed:

shell

npm install -g cline

shell

cline version

Step 1: Authenticate Cline with Kimchi

Cline manages provider configuration through its auth command:

shell

cline auth -p openai -k YOUR_CASTAI_API_KEY -b https://llm.kimchi.dev/openai/v1 -m minimax-m2.5

Flag	Value	Purpose
-p openai	openai	Tells Cline to use the OpenAI-compatible provider type
-k	Your Kimchi API key	Authenticates your requests to Kimchi
-b	https://llm.kimchi.dev/openai/v1	Points Cline at Kimchi's inference endpoint
-m	minimax-m2.5	Sets MiniMax M2.5 as the default model

Step 2: Verify Your Configuration

Confirm that Cline stored your settings correctly:

shell

cline config

You should see https://llm.kimchi.dev/openai/v1 as the openAiBaseUrl and minimax-m2.5 as the model.

Step 3: Test the Connection

Navigate to a project directory and run a simple test:

shell

cd ~/your-project

shell

cline "What files are in this directory?"

Once that succeeds, try a prompt that requires a real inference call:

shell

cline "Explain what this project does in one paragraph"

Using Cline

Cline operates in two modes:

Interactive mode — launches a full terminal interface for hands-on development sessions. Start it by running cline with no arguments.
Headless mode — designed for automation. Activates when you pass the -y flag, pipe input, or redirect output:

Run a task with automatic approval (no prompts):

shell

cline -y "Fix all ESLint errors in src/"

Pipe a diff for automated code review:

shell

git diff | cline -y "Review these changes and summarize any issues"

Advanced: Headless CI/CD Workflows

Cline's headless mode makes it straightforward to integrate Kimchi into automated pipelines:

In your CI pipeline, authenticate first:

shell

cline auth -p openai -k $CASTAI_API_KEY -b https://llm.kimchi.dev/openai/v1 -m minimax-m2.5

Then use Cline in headless mode:

shell

git diff origin/main | cline -y "Review these changes for bugs or regressions"

IDE Integration

If you use Cline as a VS Code extension, settings are stored in VS Code's settings.json:

Open VS Code and navigate to Settings (Cmd+, on macOS)
Search for Cline
Set API Provider to OpenAI Compatible
Set Base URL to https://llm.kimchi.dev/openai/v1
Set API Key to your Kimchi API key
Set Model to minimax-m2.5

Or add directly to settings.json:

json

"cline.apiProvider": "openai",
"cline.openAiBaseUrl": "https://llm.kimchi.dev/openai/v1",
"cline.openAiApiKey": "YOUR_CASTAI_API_KEY",
"cline.apiModelId": "minimax-m2.5"

📘

Settings key names may change between Cline extension versions. If the keys above don't work, check the Cline extension documentation for the current field names.