What is Kilo Code?
Kilo Code is an open-source AI coding agent for developers working in IDEs, terminals and cloud environments. Kilo Code combines code generation, feature planning, debugging and review workflows with access to hosted models, local models and provider API keys. Individual developers can start with free platform access, while teams can add collaboration and governance features through paid plans.
Kilo Code publishes MIT-licensed source so developers can inspect prompts, context windows and agent decisions. The source can also be forked, modified and self-hosted. Kilo Code supports work across VS Code, JetBrains, CLI and cloud surfaces rather than requiring every coding task to happen in a single editor.
What Kilo Code can do
Kilo Code includes specialized agents for different stages of software development. The Code Agent writes and refactors code, the Plan Agent provides structured guidance before implementation, and the Debug Agent reads errors, traces issues and suggests fixes. Cloud agents and code review extend these workflows beyond local editing.
- Kilo Code generates and refactors code with codebase context.
- Kilo Code plans complex features before code is written.
- Kilo Code traces errors and suggests bug fixes.
- Kilo Code runs local and cloud agents with parallel isolated worktrees.
- Kilo Code provides cloud-based code review workflows.
- Kilo Code routes model requests through Efficient, Frontier or Free strategies.
How Kilo Code works across editors and models
Kilo Code is available through a VS Code extension, Open VSX, a JetBrains plugin and a CLI for shells, scripts or CI jobs. Supported editor options include Cursor, Windsurf, IntelliJ, PyCharm and WebStorm. Browser-based cloud access, a Slack integration and a mobile app provide additional ways to start, follow or review agent work.
Kilo Code lets developers choose hosted inference through Kilo Gateway, supply provider keys, or run local models with Ollama or LM Studio. Supported key providers include Anthropic, OpenAI, Google, Azure and AWS Bedrock. Auto Model can select a model strategy according to the chosen capability and cost tier, including free-model routing where available.
Who Kilo Code is for
Kilo Code serves individual developers building with AI and engineering teams that need shared agents, billing and usage visibility. Enterprise plans add governance, security and support controls. Kilo Code supports both local development tasks and remote agent workflows, making the available use cases broader than code completion alone.
- Individual developers can use Kilo Code to generate, refactor and debug code.
- Developers can use Kilo Code to plan features before implementation.
- Engineering teams can use Kilo Code to share custom agents and centralize billing.
- Organizations can use Kilo Code Enterprise to restrict models or providers and access audit logs.
Kilo Code pricing and plans
Kilo Code separates platform access, AI inference and cloud compute. The individual platform plan is free, Teams costs $15/user/month, and Enterprise uses custom pricing. Teams adds shared custom agents, usage reporting, centralized billing and management features. Enterprise adds controls such as SSO, audit logs, a shared private gateway and SLA commitments.
Kilo Code can run with free models, local models or provider keys without a paid Kilo inference plan. Kilo Gateway offers pay-as-you-go inference at exact provider rates with no inference markup. Optional Kilo Pass subscriptions start at $19/mo. Credit purchases carry a 5% processing fee, and cloud compute is billed separately from inference.
Kilo Code strengths and limitations
Kilo Code combines inspectable source, multiple development surfaces and flexible model access. Developers can select hosted inference, use existing provider keys or run local models, while teams can coordinate agents and billing. These choices provide alternatives to relying on a single model provider, although free platform access does not make every usage component free.
- Kilo Code charges separately for cloud compute and model inference.
- Kilo Code applies a processing fee to credit purchases.
- Kilo Code requires paid plans for team collaboration and enterprise governance features.
- Kilo Code Desktop is offered as a beta for macOS, Windows and Linux.
Kilo Code FAQ
What is Kilo Code?
Kilo Code is an open-source AI coding agent for VS Code, JetBrains, CLI and cloud workflows. Kilo Code supports code generation, refactoring, feature planning, debugging and code review, with hosted models, provider API keys and local model options.
Is Kilo Code free?
Kilo Code offers free individual platform access and can use free models, local models or existing provider keys without a paid Kilo inference plan. Paid hosted inference and cloud compute are separate costs. Teams costs $15/user/month, while Enterprise pricing is custom.
Which editors and platforms does Kilo Code support?
Kilo Code supports VS Code, Cursor, Windsurf, Open VSX editors and JetBrains IDEs including IntelliJ, PyCharm and WebStorm. Kilo Code also offers a CLI, browser-based cloud agents, Slack access, a mobile app, and a desktop beta for macOS, Windows and Linux.
Can Kilo Code use local models or my own API keys?
Kilo Code supports local models through Ollama or LM Studio and accepts provider keys from Anthropic, OpenAI, Google, Azure, AWS Bedrock and other providers. Kilo Code also offers hosted inference through Kilo Gateway, with 500+ models across 60+ providers.
Is Kilo Code open source and self-hostable?
Kilo Code is MIT-licensed open-source software that developers can inspect, fork, modify and self-host. Kilo Code exposes prompts, context windows and agent decisions for inspection, and the website states that model switching is not performed silently.
How does Kilo Code charge for inference and cloud compute?
Kilo Code separates inference charges from cloud compute charges. Kilo Gateway bills hosted inference at exact provider rates without inference markup, while credit purchases carry a 5% processing fee. Cloud compute is usage-based, billed per second, and charged separately from model inference.
From the community
No reviews yet
Be the first to review this tool. What you write helps the next person choose.




