Kilo Code: Free Open-Source AI Coding Assistant

Kilo Code is a free, open-source AI coding agent that lives inside VS Code, JetBrains, or your terminal. Instead of just autocompleting a line here and there, it can read your whole codebase, plan a feature, write the code, run terminal commands, and fix its own mistakes. No API keys are required to get started, and once you create a free account you get access to 500+ AI models at the provider’s own price, with no markup. This guide walks you through installing Kilo Code and running your first AI-assisted coding task.

Quick Summary: Kilo Code is a free, MIT-licensed AI coding extension for VS Code and JetBrains (plus a standalone CLI). It works like an AI developer rather than a simple autocomplete tool, and you choose which underlying model powers it.

Most AI coding tools lock you into one company’s model and one price tier. Kilo Code takes the opposite approach: the extension itself is free and open source, and you plug in whichever model you actually want to pay for — GPT, Claude, Gemini, or dozens of others — at that provider’s own rate. That means your coding assistant doesn’t get more expensive or more limited just because the company behind it changes its pricing. It also means the same agent core follows you between your editor, your terminal, and even a remote server.

Where Kilo Code Works

Surface	Best For	Session Continuity
VS Code Extension	Day-to-day coding with inline diffs and a sidebar chat	Yes — resumes CLI sessions
JetBrains Plugin	IntelliJ, WebStorm, PyCharm users who want the same agent	Yes
CLI	Remote servers, SSH sessions, CI/CD pipelines	Yes — shares core with the extension
Cloud Agents	Longer background tasks and code review, without tying up your machine	Yes — resumable from any editor

Kilo Code vs. Other AI Coding Tools

Tool	Open Source?	Model Choice	Pricing
Kilo Code	Yes (MIT)	500+ models, switch anytime	Free extension, pay provider rate for models
GitHub Copilot	No	Locked to a small set of models	Flat monthly subscription
Cursor	No	Limited model selection	Flat monthly subscription

Kilo Code’s Built-In Modes

Rather than one generic chat box, Kilo Code splits work into modes so the agent behaves differently depending on what stage you’re at:

Mode	What It Does	Use It When
Architect Mode	Plans a feature and asks clarifying questions before touching code	Starting something new or complex
Code Mode	Writes and edits files directly with full diff review	You already know what needs to change
Debug Mode	Reads error output and stack traces to isolate and fix bugs	Something’s broken and you’re not sure why
Custom Subagents	Specialized agents you define for repeatable, narrow tasks	You run the same kind of task often (e.g. writing tests)

Picking a Model by Budget

Since Kilo Code doesn’t lock you into one model, it’s worth matching the model to the task instead of using your most expensive option for everything:

Task Type	Suggested Tier	Why
Quick edits, autocomplete, small fixes	Budget model (e.g. a smaller/cheaper model profile)	Low cost per call, fast responses
Everyday feature work	Mid-tier default model	Balances quality and cost for daily use
Complex refactors, architecture planning	Heavy/flagship model	Better reasoning on multi-file, high-stakes changes

Tip: Set up separate API Configuration Profiles named something like “default,” “heavy,” and “budget,” then switch between them per task instead of leaving one expensive model running for everything.

What You’ll Need

VS Code installed (or JetBrains, if you prefer that IDE)
A free Kilo Code account (only needed if you want the full 500+ model catalog)
An API key for at least one model provider (optional — some models work without one)

Step 1: Install the Extension

Step 1Open VS Code, go to the Extensions panel (Ctrl+Shift+X or Cmd+Shift+X), and search for “Kilo Code.” Click Install on the extension published by Kilo Code. If your editor can’t access the VS Code Marketplace, you can also install it from the Open VSX Registry, which supports VS Code-compatible editors like VSCodium and Windsurf.

Step 2: Sign In and Pick a Model

Step 2Once installed, open the Kilo Code panel from the sidebar and create a free account. This unlocks access to the full model catalog, including options like GPT-5.5, Claude Opus 4.7, Claude Sonnet 4.6, and Gemini 3.1 Pro Preview — all billed at the provider’s own rate. Create a couple of API Configuration Profiles if you want to switch quickly between a cheaper model for routine work and a stronger one for harder tasks.

Warning: You are billed directly by the model provider based on usage, not by Kilo Code itself. Keep an eye on the built-in cost tracker in the status bar so a long agent session doesn’t run up an unexpected bill.

Step 3: Choose a Mode

Step 3Kilo Code ships with specialized modes for different parts of the workflow. Architect Mode helps you plan a feature before any code is written, while regular coding mode writes and edits files directly. Pick the mode that matches what you’re trying to do before you type your first prompt.

Step 4: Give It a Task

Step 4Type a plain-language description of what you want in the chat panel — for example, “add form validation to the signup page and write a test for it.” Kilo Code will read the relevant files in your project, propose a plan, and start editing. You’ll see a diff for every change before it’s applied.

Step 5: Review and Approve Changes

Step 5Use the diff panel to review each change Kilo Code makes. You can accept, reject, or ask for revisions on individual edits rather than approving the whole task blindly. This keeps you in control even when the agent is working across multiple files at once.

Step 6: Run and Debug

Step 6Kilo Code can run terminal commands on your behalf, including your test suite. If something fails, you can hand the error output straight back to the agent and ask it to fix the failure — it will read the stack trace and attempt a correction automatically.

Tip: You can start a task in your terminal over SSH and reopen the same session later inside VS Code, since the extension and CLI share the same underlying core. That makes it easy to kick off a long task on a remote server and check back later.

The Kilo Code Workflow at a Glance

Install Extension → Sign In → Choose a Model → Pick a Mode (Architect / Code) → Describe the Task → Review the Diff → Approve or Revise → Run & Debug

Example Prompts to Try

“refactor this function to remove duplicate code and add comments”
“write unit tests for the auth module”
“find and fix the bug causing the 500 error on checkout”
“plan out a REST API for a simple to-do list app”

Troubleshooting Common Issues

Extension won’t install from the Marketplace: If your organization blocks the VS Code Marketplace, install Kilo Code from the Open VSX Registry instead — it works the same way in VS Code-compatible editors like VSCodium and Windsurf.

401 / Unauthorized error: Double-check your API key was pasted without extra spaces, and regenerate it from your provider’s dashboard if the error persists.

404 / Model not found: Confirm the exact model ID you selected is actually available on your account — some models require separate access approval from the provider.

Agent runs but on a managed/corporate network fails to connect: Set your organization’s proxy in VS Code’s network settings rather than in a separate CLI install, and make sure your org’s certificate authority is trusted by your OS if HTTPS inspection is in use.

Frequently Asked Questions

Is Kilo Code really free? The extension itself is free and open source under the MIT license. You only pay the model provider directly for the AI usage, at their standard rate with no added markup.

Do I need an API key to start? No — you can begin without one, though creating a free account and adding a provider API key gives you access to the full range of 500+ models.

Does it work outside VS Code? Yes. Kilo Code also has a JetBrains plugin and a standalone CLI that shares the same session core, so you can start a task in one place and pick it up in another.