Kilo Code: The Open Source AI Coding Tool (2026)

Kilo Code is an open source AI coding assistant that runs as a VS Code extension. It is a fork of Cline (itself originally called Claude Dev) with a focus on improved UX, lower token usage, and better support for multiple AI providers. If you have been looking for a free, open source alternative to Cursor's AI features that works inside VS Code, Kilo Code is worth evaluating.

What Is Kilo Code?

Kilo Code is a VS Code extension that gives you an AI coding agent directly in your editor. It can:

Read and write files in your project
Execute terminal commands
Browse the web for documentation
Create and edit multiple files in a single task
Use any AI model via API (Claude, GPT, Gemini, local models)
Show you exactly what it plans to do before executing

It differs from basic code completion tools (like Copilot) because it is an agent -- it can perform multi-step tasks autonomously, like "create a new REST API endpoint with tests and documentation."

Kilo Code vs Cline vs Cursor

Feature	Kilo Code	Cline	Cursor
Open source	Yes (Apache 2.0)	Yes (Apache 2.0)	No
Runs in VS Code	Yes	Yes	Separate editor (VS Code fork)
Model support	Any API (Claude, GPT, Gemini, Ollama)	Any API	Claude, GPT (built-in)
File editing	Yes (with approval)	Yes (with approval)	Yes (inline + chat)
Terminal access	Yes	Yes	Yes
Web browsing	Yes	Yes	No
Token optimization	Improved over Cline	Baseline	Proprietary
Price	Free (bring your API key)	Free (bring your API key)	$20/month + API costs
Custom modes	Yes	Limited	No

Installation

From VS Code Marketplace

1. Open VS Code
2. Go to Extensions (Ctrl+Shift+X / Cmd+Shift+X)
3. Search for "Kilo Code"
4. Click Install
5. Reload VS Code if prompted

From the Command Line

code --install-extension kilocode.kilo-code

From Source (for contributors)

git clone https://github.com/kilocode/kilo-code.git
cd kilo-code
npm install
npm run build

# Package the extension
npx vsce package

# Install the .vsix file
code --install-extension kilo-code-*.vsix

Configuration

After installing, configure your AI provider.

Step 1: Open Kilo Code Settings

Press Ctrl+Shift+P (or Cmd+Shift+P on macOS) and type "Kilo Code: Open Settings".

Step 2: Configure an AI Provider

Kilo Code supports multiple providers. Here are the most common:

Anthropic (Claude):

{
  "kilocode.provider": "anthropic",
  "kilocode.apiKey": "sk-ant-xxxxx",
  "kilocode.model": "claude-sonnet-4-20250514"
}

OpenAI (GPT):

{
  "kilocode.provider": "openai",
  "kilocode.apiKey": "sk-xxxxx",
  "kilocode.model": "gpt-4o"
}

Google (Gemini):

{
  "kilocode.provider": "google",
  "kilocode.apiKey": "AIzaSyxxxxx",
  "kilocode.model": "gemini-2.0-flash"
}

Local models (Ollama):

{
  "kilocode.provider": "ollama",
  "kilocode.baseUrl": "http://localhost:11434",
  "kilocode.model": "qwen2.5-coder:32b"
}

OpenRouter (access all models):

{
  "kilocode.provider": "openrouter",
  "kilocode.apiKey": "sk-or-xxxxx",
  "kilocode.model": "anthropic/claude-sonnet-4-20250514"
}

Using Kilo Code

Basic Usage

Open the Kilo Code panel from the activity bar (the Kilo icon on the left sidebar) or press the keyboard shortcut. Then type your request in natural language:

Create a Node.js Express server with a /health endpoint
that returns the current timestamp and uptime.

Kilo Code will:

Plan the file changes
Show you exactly what it will create or modify
Wait for your approval
Execute the changes
Optionally run commands (like npm install express)

Multi-File Tasks

Kilo Code excels at tasks that span multiple files:

Add user authentication to this Express app:
- Create a User model with email and hashed password
- Add /register and /login endpoints
- Create JWT middleware for protected routes
- Add tests for the auth endpoints

Terminal Commands

Kilo Code can run terminal commands with your approval:

Install the required dependencies and run the test suite

It will show you the exact commands before running them:

# Kilo Code proposes:
npm install bcryptjs jsonwebtoken
npm test

You can approve, modify, or reject each command.

Web Browsing

Kilo Code can fetch documentation from the web to inform its work:

Look up the latest Prisma ORM migration syntax and create a
migration for adding a "comments" table with foreign keys to
the "users" and "posts" tables.

Custom Modes

One of Kilo Code's standout features is custom modes. Modes let you configure the AI's behavior for different types of work:

Built-in Modes

Mode	Description	Tool Access
Code	Default coding mode	Full file, terminal, browser access
Architect	Planning mode	Read-only file access, no terminal
Ask	Q&A mode	Read-only, informational responses
Debug	Debugging mode	Full access with diagnostic focus

Creating a Custom Mode

Create a .kilo/modes.json file in your project root:

{
  "modes": [
    {
      "slug": "reviewer",
      "name": "Code Reviewer",
      "roleDefinition": "You are a senior code reviewer. Analyze code for bugs, security issues, and best practices. Never modify files directly.",
      "groups": ["read"],
      "customInstructions": "Focus on: error handling, input validation, SQL injection, XSS, and race conditions. Be specific about line numbers."
    },
    {
      "slug": "docs",
      "name": "Documentation Writer",
      "roleDefinition": "You are a technical writer. Generate clear, concise documentation.",
      "groups": ["read", "edit"],
      "customInstructions": "Write JSDoc comments for functions, README sections, and inline comments for complex logic."
    }
  ]
}

Switch modes using the dropdown in the Kilo Code panel or by typing /mode reviewer.

Token Usage Optimization

Kilo Code includes several features to reduce token consumption:

Context Management

# Kilo Code automatically:
- Sends only relevant files to the AI (not the entire project)
- Truncates large files to relevant sections
- Caches file reads within a conversation
- Uses diff-based editing to minimize output tokens

Manual Context Control

You can explicitly control what context is sent:

@file:src/auth/middleware.ts @file:src/models/user.ts
Fix the JWT verification in the auth middleware - it should
check token expiration properly.

The @file: prefix tells Kilo Code to include specific files in the context.

Cost Tracking

Kilo Code shows real-time token usage and cost estimates in the panel:

Task: Add user authentication
Tokens used: 12,450 input / 3,200 output
Estimated cost: $0.08 (Claude Sonnet 4)

Configuration File Reference

Global settings in VS Code settings.json:

{
  "kilocode.provider": "anthropic",
  "kilocode.model": "claude-sonnet-4-20250514",
  "kilocode.apiKey": "sk-ant-xxxxx",
  "kilocode.maxTokens": 8192,
  "kilocode.temperature": 0,
  "kilocode.autoApprove": {
    "readFiles": true,
    "listFiles": true,
    "writeFiles": false,
    "executeCommands": false
  },
  "kilocode.customInstructions": "Always use TypeScript. Prefer functional components in React. Use Tailwind CSS for styling."
}

Per-Project Instructions

Create a .kilo/instructions.md file in your project root for project-specific instructions:

# Project Instructions

- This is a Next.js 15 app with the App Router
- Use Prisma for database access
- All API routes should validate input with Zod
- Tests use Vitest
- Use pnpm, not npm or yarn

Troubleshooting

Issue	Cause	Solution
"API key invalid"	Wrong key or expired	Regenerate key from provider dashboard
Extension not loading	VS Code version too old	Update VS Code to latest version
Slow responses	Large context or slow model	Use a faster model; reduce context with `@file:`
File edits not applied	Permission issue	Check workspace trust settings in VS Code
Terminal commands fail	Shell not detected	Set `kilocode.shell` in settings
High token usage	Sending too much context	Use `@file:` to limit context; close irrelevant files

Conclusion

Kilo Code brings autonomous AI coding agent capabilities to VS Code without requiring a subscription or a separate editor. Its support for multiple AI providers, custom modes, and token optimization makes it a strong open source alternative to proprietary tools. The key advantage is flexibility: you choose your AI provider, control your costs, and customize the agent's behavior for your workflow.

For developers using Kilo Code to build applications that involve AI-generated media -- video creation, avatar generation, or image processing -- Hypereal AI provides a simple, affordable API for integrating these capabilities into your projects without managing GPU infrastructure.