Multi-surface agent
Use the same agent across VS Code, JetBrains, CLI, and cloud workflows so work can move between surfaces without changing tools.
Kilo
ПодтвердитьKilo is an open source AI coding agent for VS Code, JetBrains, CLI, and cloud workflows. It supports model routing, bring-your-own-key and local model setups, and hosted agent workflows through KiloClaw.
Open source AI coding agent for IDE, CLI, and cloud
Kilo is an open source AI coding agent built for development workflows across IDEs, the CLI, and cloud agents. The site positions it as a single agent that can follow work across VS Code, JetBrains, command-line sessions, and hosted environments.
Its core purpose is to help developers code, refactor, debug, plan, and review with model choice left open: users can rely on Kilo’s Auto Model routing, bring their own provider keys, or use local models. The product also includes Kilo Gateway for model inference access and KiloClaw for hosted OpenClaw agents that keep working after the IDE closes.
Core features
Specialized agent modes
Switch between Code, Architect, Debug, Ask, and Custom modes to match the task, from implementation to planning and troubleshooting.
Automatic model routing
Let Auto Model route each session to Efficient, Frontier, Balanced, or Free based on task type and cost/capability goals.
Broad model support
Connect to 500+ hosted models and use provider keys from services such as Anthropic, OpenAI, Google, Azure, or AWS Bedrock, or run local models with Ollama or LM Studio.
Gateway API
Use Kilo Gateway as a universal AI inference API for routing LLM requests or powering custom apps with the same provider access model.
Hosted agent workflows
Start hosted work in KiloClaw for always-on OpenClaw agents with chat channels, scheduled automations, browser control, and remote execution support.
Common use cases
Day-to-day coding in the IDE
Use Kilo Code inside an editor to write new code, refactor existing files, and keep the agent working within the project context.
Planning and troubleshooting
Move into Architect or Debug mode when you need a plan for a larger feature, or when you want help tracing and fixing an error.
Cross-surface agent workflows
Use the CLI or cloud agents when the work should continue outside the IDE, including remote execution and always-on tasks.
Always-on hosted agents
Let a hosted KiloClaw agent continue after the IDE closes for chat-based coordination, scheduled automations, or long-running jobs.
Model access and routing
Route requests through Kilo Gateway or Auto Model when you want provider access, model routing, or cost control without manually choosing a model each time.
Pros and Cons
Pros
- Works across VS Code, JetBrains, CLI, and cloud agents.
- Supports 500+ models, plus bring-your-own-key and local model workflows.
- Auto Model reduces manual model switching by routing sessions to a suitable model strategy.
- Open source design exposes prompt, context, and decisions in MIT-licensed source.
- Offers both local coding workflows and hosted always-on agent workflows through KiloClaw.
Cons
- Some capabilities are distributed across separate products and tiers, such as Kilo Code, Kilo Gateway, Kilo Pass, and KiloClaw, so users may need to choose the right setup for their workflow.
- The source highlights many supported surfaces and model options, but the documentation shown here is only partial on details like team workflows, security controls, and exact integration behavior.
FAQ
What is Kilo Code?
Kilo Code is the open source AI coding agent for IDE, CLI, and cloud workflows. It is designed to help developers write, refactor, debug, and review code across supported environments.
Is Kilo Code free?
Yes. The site describes a free and open source Kilo Code plan, with AI inference billed separately. It also offers other pricing options such as Kilo Gateway pay-as-you-go, Kilo Pass, Teams, Enterprise, and KiloClaw.
Which platforms does Kilo Code support?
Kilo Code is available for VS Code, JetBrains IDEs, the CLI, and cloud agents. The install page also highlights setup in about 60 seconds for VS Code, JetBrains, and CLI.
How does model selection work?
Kilo Code uses Auto Model to route requests to different model strategies, including Efficient, Frontier, Balanced, and Free. Users can still choose hosted models, bring their own provider keys, or run local models with Ollama or LM Studio.
How does Kilo Code handle transparency?
The site says Kilo Code is open source and lets users inspect the prompt, context window, and decisions in MIT-licensed source. It also says there is no silent model switching.
Quick Facts
- Category
- Developer Tool
- Source domain
- kilocode.ai
- Primary use
- AI-assisted coding, debugging, planning, and review
- Supported surfaces
- VS Code, JetBrains, CLI, cloud agents
- Model access
- 500+ hosted models, BYOK, and local models
- Pricing shape
- Free open source product with separate AI inference pricing options