Kilo

Open-source AI coding agent, covering VS Code, JetBrains, CLI and cloud, supporting access to more than 500 models and zero markup for self-provided keys, and using local models can protect code privacy.

AI Developer Tools#Open Source#Coding Agent#Multi-Model

Updated: 2026-10-26

What is Kilo?

Kilo is an open-source AI coding agent. Its official website highlights the core selling point in a straightforward way: "Write code with AI without emptying your wallet." It targets both model selection flexibility and cross-tool availability. The official claims it supports integration with over 500 models, comes with built-in keys with no additional charges, and covers four forms: VS Code, JetBrains, CLI, and cloud. This single agent can be used in almost all work scenarios you can think of. Official data shows it has over 5 million users, processes more than 10 trillion tokens per month, and has been named the Best Open-Source Product of the Month. The official website also continuously releases content related to model selection and security evaluations.

Who is it for?

Independent developers and small teams sensitive to model costs and choices; engineering organizations that need a unified coding experience across multiple editors and cloud environments; developers who want to deeply understand the implementation of AI coding agents or perform custom development on them. It is also suitable for leaders who wish to unify their team's AI coding tools and reduce decentralized procurement, covering different roles with a single entry point.

Key features

500+ models to choose from

You can use cutting-edge closed-source models and open-weight models, and you can also enter your own API keys. The choice of models is fully left to the user, and there are corresponding options whether you pursue cutting-edge effects or cost control.

No Extra Markup

The official states that models are provided at cost, with no price increases and no vendor lock-in, and this is a key focus of its promotion.

Available on all platforms

Provides VS Code extensions, JetBrains plugins, command-line and cloud versions, ensuring a consistent experience in editors. The cloud version can handle heavy tasks in succession, so you don’t need to change your habits when switching tools.

Local Model for Privacy Protection

Local models can be used to process code, preventing code from being leaked externally and meeting scenarios with strict data boundary requirements.

Open-source and Auditable

The official emphasizes "No black boxes, inspect everything," achieving full open-source and traceability. It also provides an Agent Command Center to manage multiple agents through a single entry point, and releases evaluation information such as model security scores to assist in selection.

How to use Kilo

The key differences between such "multi-model aggregation" tools lie in billing transparency and stability.
It is recommended to first confirm the billing method and usage limits for self-provided keys.
You can start with the VS Code extension, run a real task, and then consider expanding to JetBrains and CLI to gradually build a unified usage method for the team.
pay attention to the official released model evaluations and safety scores, which can assist in selection decisions.

Things to know

When using third-party cloud models, code will be sent to the corresponding model service providers; for sensitive projects, it is recommended to switch to local models instead.

The differences in features and quotas between the free and paid versions are subject to the official website; confirm your usage tier before integration.

Integrating multiple models means you need to keep an eye on the version, rate limiting, and billing strategies of each model to avoid cost out of control.

The open-source project is updated rapidly. It is recommended to pay attention to the version change notes before upgrading to avoid disrupting the workflow; the usage and billing methods of the cloud version are subject to the official website, and you should confirm the account system before accessing it.