Roo Code

Open-source VS Code coding agent with switchable modes and your own key

AI Coding & DevelopmentFree planOverseasβ˜…β˜…β˜…β˜…β˜† 4.0

What is Roo Code?

Roo Code is an open-source VS Code extension that puts an autonomous coding agent inside the editor. What sets it apart is a mode system: separate profiles such as Code, Architect, Ask and Debug each carry their own instructions and permitted tools, so planning, implementing and answering questions happen under different rules rather than through one catch-all prompt. It supports many model providers as well as local models through Ollama or LM Studio, and custom modes can be stored as files in the repository and shared with a team. You bring your own API key, so the extension itself carries no subscription.

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

Key features

  • Autonomous multi-file editing inside VS Code with reviewable diffs
  • Switchable modes with separate instructions and tool permissions
  • Bring your own key or connect a local model runtime
  • Custom modes saved as files that a team can share
  • Terminal command execution behind approval prompts
  • Open source under an Apache licence with community extensions

Pros & cons

Strengths

  • No subscription; you pay only for the model usage you choose
  • Mode separation keeps risky tools out of casual conversations
  • Local models are supported for privacy-sensitive repositories

Watch out for

  • Token consumption can climb quickly on large tasks
  • Setup and model selection are left almost entirely to you
  • Agent quality varies a lot by model and mode configuration

Best for & use cases

developers, self-hosted setups, local model users and teams wanting agent control

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

FAQ

How is it different from other VS Code agents?

The mode system is the main difference. Instead of one agent with one prompt, you configure separate profiles with their own instructions and tool access, which makes planning and executing behave more predictably.

Do I need a paid API key?

Not necessarily. You can connect a local model through Ollama or LM Studio and avoid API costs, though a strong hosted model generally performs better on multi-step tasks.

Is it safe to let it run commands?

It asks for approval before executing terminal commands by default, and auto-approval is opt-in. Leaving approval on is advisable until you trust its behaviour in a given repository.