Kilo Code

Open-source VS Code agent that can orchestrate other coding agents

AI Coding & DevelopmentFree planOverseasβ˜…β˜…β˜…β˜†β˜† 3.0

What is Kilo Code?

Kilo Code is an open-source VS Code extension that combines autonomous coding with a deliberate emphasis on control. It offers an architect mode for planning, a code mode for implementation and a debug mode for investigating failures, along with an orchestration mode that breaks a large task into parts and delegates them. You bring your own model key or connect a local runtime, and a wide range of providers is supported. It is a younger project than the category leaders, which shows in polish rather than ambition, and the source is fully inspectable.

Last updated: 2026-09-20. This site only provides an index; for exact features, pricing, and licensing, see the official website.

Key features

  • Autonomous coding inside VS Code with reviewable diffs
  • Separate architect, code and debug modes with distinct prompts
  • Orchestrator mode that splits a task and delegates the parts
  • Bring your own API key or connect a local model runtime
  • Model Context Protocol support for custom tools and servers
  • Open source under an Apache licence with active development

Pros & cons

Strengths

  • No subscription beyond the model usage you choose
  • Orchestration helps on longer multi-step tasks
  • Fully open source, so its behaviour can be inspected

Watch out for

  • Less polished than the established commercial agents
  • Token costs are easy to underestimate on large tasks
  • Frequent releases mean the occasional rough edge

Best for & use cases

developers, cost-conscious agent use, local models and self-hosted workflows

If you're comparing similar products, check the alternatives below, or browse all tools in the AI Coding & Development category.

FAQ

What is orchestration mode for?

It lets one agent sequence a larger piece of work by splitting it into subtasks with dedicated roles. It helps on multi-step features, though the results still need review like any agent output.

How does it handle my code?

Code goes to whichever model provider you configure, or stays on your machine if you run a local model. That choice is explicit, which matters when working under data-handling rules.

Is it ready for daily use?

Many developers run it daily and releases come quickly. Expect occasional rough edges, and pin your extension version if stability matters more to you than new features.