Kilo Code AI Agent: Open Source Coding Review
/kilocode
Kilo Code is an open source AI coding agent that plugs into VS Code, JetBrains, and the command line, letting you pick from 500+ models and switch between them mid-task at the provider's own rate. Reach for it if you want one agent across editor, CLI, and CI without a subscription markup; skip it if you'd rather stick to your editor's built-in assistant and not manage API keys or model choice yourself.
Understanding Kilo Code
Kilo Code is a TypeScript-based, MIT-licensed coding agent you install as a VS Code extension, a JetBrains plugin, or a CLI binary. It ships five built-in agent modes — Code, Plan, Ask, Debug, and Review — each suited to a different point in the workflow, and it connects to whichever LLM provider you choose rather than routing everything through one built-in model.
Key Features of Kilo Code
- ✓Five specialized agent modes — Code for implementing and editing, Plan for architecture and implementation plans before code is written, Ask for read-only codebase questions, Debug for tracing issues, and Review for scanning changes across performance, security, style, and test coverage.
- ✓Choice of 500+ models with mid-task switching, billed at the provider's own rate with no markup, and no API key required just to start.
- ✓Inline autocomplete functionality, featuring ghost-text suggestions that can be accepted by pressing tab, alongside multi-file code generation from natural language.
- ✓Self-checking: the agent reviews and corrects its own output before handing it back.
- ✓Terminal and browser control, so the agent can run commands and automate web tasks directly instead of only editing files.
- ✓An MCP marketplace for finding and wiring up MCP servers that extend what the agent can call.
- ✓Autonomous Mode (`kilo run --auto`) that skips all permission prompts for unattended runs in CI/CD pipelines.
- ✓Custom agents you can build on top of the five built-in ones.
Practical Applications for Kilo Code
- •Editing and generating code across multiple files inside VS Code or a JetBrains IDE without leaving the editor.
- •Running an unattended fix in a CI/CD pipeline with `kilo run --auto "run tests and fix any failures"`.
- •Setting up automated AI code reviews on pull requests through app.kilo.ai/code-reviews.
- •Working from the CLI or a cloud-hosted agent at app.kilo.ai/cloud when you don't want to run anything locally.
- •Extending the agent's reach with MCP servers pulled from the MCP marketplace.
Installing Kilo Code
Kilo Code installs differently depending on where you want to run it. In VS Code, install the Kilo Code extension directly or from the VS Code Marketplace, then create an account to get model access. In JetBrains IDEs, install the Kilo Code plugin from the JetBrains Marketplace or search "Kilo Code" under Settings → Plugins. For the CLI, install with `npm install -g @kilocode/cli`, `pnpm add -g @kilocode/cli`, `bun add -g @kilocode/cli`, `curl -fsSL https://kilo.ai/cli/install | bash`, `brew install Kilo-Org/tap/kilo` on macOS/Linux, or `paru -S kilo-bin` on Arch, then run `kilo` in a project directory. Prebuilt binaries for Windows, macOS (Apple Silicon and Intel), and Linux (x64 and ARM) are also on the GitHub Releases page, and a `musl` build covers Alpine or minimal Docker images without glibc.
Using Kilo Code's AI Agents
Kilo Code defaults to the Code agent for implementing and editing code from natural-language instructions. Switch to Plan when you want an architecture and implementation plan written before any code changes happen, to Ask for questions about the codebase that don't touch files, to Debug for tracing an issue, or to Review to have the agent scan a change for performance, security, style, and test-coverage problems. For unattended runs, `kilo run --auto "run tests and fix any failures"` executes without any permission prompts — the README is explicit that this should only run in trusted environments, since `--auto` lets the agent execute any action without confirmation.
Strengths
- ✓One agent works across VS Code, JetBrains, and the CLI, so you're not locked into a single editor's assistant.
- ✓Model choice isn't fixed: pick from 500+ models and switch mid-task, paying the provider's own rate with no markup.
- ✓Five distinct agent modes map onto real workflow stages instead of one do-everything chat box.
- ✓MIT license and open source, so the code itself is inspectable and modifiable.
- ✓Autonomous Mode gives you a real CI/CD hook (`kilo run --auto`) instead of only an interactive chat loop.
Important Considerations for Kilo Code
- △`--auto` disables every permission prompt, so a CI/CD run can execute arbitrary commands with no human confirmation — the README itself warns to only use it in trusted environments.
- △Full model access requires creating a Kilo account, even though the tool is open source; the README only says "no API keys required to start," not that the whole model catalog is available account-free.
- △JetBrains and CLI users get the same five agent modes as VS Code, but the README's IDE-integration detail (ghost-text autocomplete, etc.) is written from the VS Code angle, so how complete that surface is on JetBrains isn't spelled out.
- △As a CLI built on a fork of OpenCode and bundled into a larger platform, cloud agent, code review, and KiloClaw features depend on Kilo's own infrastructure (app.kilo.ai), not just a standalone binary you run offline.
Comparing Kilo Code with Other Tools
Frequently Asked Questions
Kilo Code is an open source AI coding agent that works inside VS Code, JetBrains IDEs, and the command line, letting you pick from 500+ models instead of being locked to one.
Install the Kilo Code extension from the VS Code Marketplace, the Kilo Code plugin from the JetBrains Marketplace, or the CLI with `npm install -g @kilocode/cli` (curl, pnpm, bun, Homebrew, and AUR options are also available).
Kilo Code supports 500+ models that you can switch between mid-task, including options like GPT-5.5, Claude Opus 4.7, Claude Sonnet 4.6, and Gemini 3.1 Pro Preview, billed at the provider's own rate.
Kilo Code itself is open source and free under the MIT license; you pay the model provider's rate directly with no markup from Kilo, and no API keys are required just to start.
Kilo Code supports automated code reviews on pull requests, set up at app.kilo.ai/code-reviews, and a dedicated Review agent scans changes for performance, security, style, and test-coverage issues.
Autonomous Mode is Kilo Code's unattended workflow, triggered with `kilo run --auto`, that skips all permission prompts so the agent can execute actions without confirmation — built for CI/CD pipelines but only recommended in trusted environments.
The problem it solves
Developers already running agentic coding tools tend to get locked to one editor and one model vendor — GitHub Copilot ties you to its own models inside VS Code, and most agent CLIs work in only one environment. Kilo Code's problem statement is narrower: give the same agent, the same five modes, and the same model catalog whether you're in VS Code, JetBrains, the terminal, or a CI pipeline, instead of maintaining a separate tool per surface.
Who should try it — and who should skip
Reach for Kilo Code if you already juggle multiple LLM providers and want one agent that follows you between VS Code, JetBrains, and CI without a subscription markup. Skip it if you're happy with your editor's built-in assistant and don't want to manage a separate account and model picker for coding tasks.
Related repositories
Still deciding about kilocode?
One click hands the question to an AI along with this page — see what it says about kilocode.