Guide · Open source
How to use Kilo Code: install the CLI, VS Code or JetBrains extension
Updated · written with AI from the project's own documentation and checked against our catalogue data
Kilo Code is an open source AI coding agent that works inside VS Code, JetBrains IDEs, or a terminal CLI. You describe a change in plain language, and the agent writes or edits code across multiple files, can run commands, and can switch between many AI models. This guide shows you how to install it, pick an agent, and run it safely.
Quick start
- 1Pick your environment. Choose VS Code, a JetBrains IDE, or the terminal CLI depending on where you work.
- 2Install the tool. Install the extension or plugin from its marketplace, or install the CLI with npm, pnpm, bun, Homebrew, AUR, or the curl script.
- 3Open a project. In the editor, open your project. In the terminal, cd into the project directory.
- 4Start Kilo. Run kilo in the project directory to start the agent.
- 5Pick an agent. Use Code for edits, Plan for designs, Ask for questions, and Debug for troubleshooting.
- 6Describe the change. Explain the change in plain language and review the files the agent writes or edits.
- 7Automate carefully. For CI/CD, use kilo run with --auto only in trusted environments and with rules that deny unwanted actions.
Install Kilo Code in VS Code, JetBrains or the terminal
Choose the environment you work in. For editor use, install the Kilo Code extension for VS Code from the VS Code Marketplace, or the Kilo Code plugin for a JetBrains IDE from the JetBrains Marketplace. In a JetBrains IDE you can also search for "Kilo Code" under Settings → Plugins.
For terminal use, install the CLI with npm, pnpm, bun, Homebrew or Arch Linux's AUR, or with the curl script. The npm-based install needs Node.js or a package manager such as npm, pnpm, or bun.
- VS Code or a JetBrains IDE, for editor use
- Node.js or npm, pnpm, or bun, for the npm-based CLI install
npm install -g @kilocode/cliStart Kilo in a project directory
After installing the CLI, change into the project you want to work on and run kilo. The source says to run kilo in any project directory to start.
Create an account to get access to the models listed for the extension. The README says you can start without an API key, and you pay the model provider's rate with no markup.
cd your-project
kiloChoose the right agent for the task
Kilo ships with specialized agents that you switch between depending on what you need. Code is the default and implements or edits code from natural language. Plan designs architecture and writes implementation plans before any code is written. Ask answers questions about your codebase without touching files. Debug troubleshoots and traces issues.
You can also build your own custom agents. The README links to the documentation on agents and custom agents.
- Code: the default, implements and edits code
- Plan: designs architecture and writes plans first
- Ask: answers questions without changing files
- Debug: troubleshoots and traces issues
Use autonomous mode in CI/CD pipelines
For pipelines, run kilo run with the --auto flag. The README describes this as fully autonomous operation with no prompts, built for CI/CD.
Be careful with this mode. --auto auto-approves permission prompts unless a rule explicitly denies the action, so use it only in trusted environments and set rules that deny what you do not want.
- Write rules that deny actions you do not want before enabling --auto
- Use it only in trusted environments
kilo run --auto "run tests and fix any failures"Understand what Kilo can do
Kilo generates code from natural language across multiple files, and it can run terminal commands and control a browser. Its self-checking step lets the agent review and correct its own work. Inline autocomplete shows ghost-text suggestions that you accept with Tab.
The Kilo Marketplace lets you install agents, skills, MCP servers, and plugins that extend what the agent can do. You can switch between 500+ models mid-task to match latency, cost, and reasoning to the job.
- Code generation across multiple files
- Inline autocomplete with Tab to accept
- Terminal and browser control
- Kilo Marketplace for agents, skills, MCP servers and plugins
Find documentation and contribute
For configuration and other details, the README points to the official docs at kilo.ai/docs. If you want to contribute, start with the Contributing Guide, which covers environment setup, coding standards, and how to open a pull request. RELEASING.md covers the VS Code extension and CLI release process, and packages/kilo-jetbrains/RELEASING.md covers the JetBrains plugin.
What people use it for
- Describe a feature in plain language and have the agent edit several files in a VS Code project
- Ask the Ask agent questions about an unfamiliar codebase without changing any files
- Use the Plan agent to outline an architecture before writing code
- Run kilo run --auto in a CI/CD pipeline to run tests and fix failures in a trusted environment
- Switch between different AI models for different tasks inside one session
Frequently asked questions
Is Kilo Code free?
Kilo Code is MIT licensed, so you can use, modify, and distribute the code, including commercially, as long as you keep the attribution and license notices. The README says you pay the model provider's rate with no markup, and you can start without an API key. Check the provider's pricing for the models you choose.
What do I need to install Kilo Code?
For the CLI, the README lists Node.js or a package manager such as npm, pnpm, or bun. For editor use, you need VS Code or a JetBrains IDE. The CLI can also be installed from GitHub Releases as a binary.
How do I install the Kilo CLI?
Run npm install -g @kilocode/cli, or use the curl install script, pnpm add -g @kilocode/cli, bun add -g @kilocode/cli, Homebrew, or AUR. Then run kilo in any project directory to start.
Can Kilo Code run offline?
No. The README describes use of model providers and a cloud service, so Kilo is not a fully offline tool.
Is Kilo Code safe to run in automation?
Use --auto only in trusted environments. The README says it auto-approves permission prompts unless a rule explicitly denies the action, so define deny rules first.
How does Kilo Code compare to Cline, Aider, OpenCode or Continue?
Kilo Code works in VS Code, JetBrains, and a CLI, and offers many models and a marketplace. The README notes that its CLI is a fork of OpenCode. Other tools in the same space include Cline, Aider, and Continue, so compare the features you need before choosing.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Kilo Code suits developers who want an open source AI coding agent in VS Code or JetBrains, want to choose among many models, or want to automate checks in CI/CD. It is not a good fit for people who need a fully offline tool or who do not write code, and it should not be run unattended without carefully written deny rules.
Install options & downloads