Kilo Code is an AI coding agent that runs in several places: as an extension for VS Code, in JetBrains IDEs such as IntelliJ IDEA, PyCharm and WebStorm, as a command-line tool, as cloud agents, and inside Slack. The project is MIT licensed and open source, and its site says you can fork it, modify it and self-host it.
Its pitch is choice and transparency about models. The site advertises 500+ models, covering frontier models, open-weight models and your own keys, with local models supported. It says there is no silent model switching and that prompts and context are visible to you. Work is organized into modes: Code mode to write and refactor code, Architect mode to plan larger features, and Debug mode to find and fix bugs. There is also an Auto Model option that routes between Efficient, Frontier and Free tiers so you do not have to choose each time.
On price, the site offers a free way to start with no credit card required, and describes usage as token-based with zero markup on AI inference and no lock-in. In practice that means your bill follows how many tokens you use and which models you pick, which is fair but less predictable than a flat subscription. The page we checked does not spell out the free tier’s limits, so confirm those before you rely on them.
Verdict: a strong option if you want an open agent, a wide menu of models and pay-as-you-go pricing. It sits in the same space as Cline, and if you want the editor itself to be the AI product, compare Cursor. The best AI coding tools guide covers the rest.
Sources: Kilo Code’s website, checked on September 21, 2026.
