📐

roo tools in VS code

Roo Code is an AI-powered autonomous coding agent embedded directly in your VS Code editor which you can download for free from Microsoft and just uses AI model tokens you provide.

It can read/write files, run terminal commands, control a browser (with certain models), and can work off natural language instructions.

It is highly customisable, model agnostic, and packed with productivity tools like multi mode workflows and smart automation protocols.

Its key features

  • You can configure it to run in a ‘smart mode’ - which is like a filter on how it will behave in response to questions. The modes include:
    • Code mode:
      • Everyday coding, generation, and refactors.
    • Architect mode:
      • High-level design, system thinking, and architecture planning.
    • Ask mode:
      • Query your codebase or technical concepts in depth.
    • Debug mode and orchestrator mode:
      • For systematic diagnosing and multi-step task management.
    • Custom Modes:
      • Create tailored personas (e.g., QA engineer, product manager).
  • File and editor operations
    • Create/edit files with diff previews.
    • Auto-detect and fix compiler or linter errors.
    • Supports drag-and-drop file interactions.
  • Terminal and command integration
    • Run tasks:
      • build, test, install packages.
    • Live feedback:
      • Roo reacts to terminal output, errors, or logs.
    • Supports "Proceed While Running" for long-running processes.
  • Browser automation
    • Launch web apps, click buttons, capture screenshots/logs.
    • Ideal for e2e testing, UI debugging, or reproducing issues.
  • Model-agnostic support
    • Compatible with OpenAI, Gemini, Claude, Anthropic, DeepSeek, AWS, local LLMs, etc.
    • API usage tracked with detailed cost and token stats.
  • Advanced customisation via MCP
    • Integrate external APIs or tools using the Model Context Protocol (MCP).
    • Add capabilities: database queries, AWS management, Jira tickets, PagerDuty calls, and more.
  • Context mentions
    • Use @file, @folder, @git, @url, @problems to direct Roo to specific context.
  • Auto-approval and iteractive modes
    • Set auto-approve rules (e.g., for diffs or browser actions).
    • Modes can auto-switch based on the task, offering hands-free automation. \

A lot of the above sounds technical - but what the tool is essentially doing is giving you crazy levels of power and control to complete tasks conversationally. You still need to know conceptually what you’re doing, but the technical barrier to entry to achieve quite complex things is probably 1/100 of what is required via a traditional path.

Popular use cases

  • Rapid prototyping: Generate boilerplate, build features, scaffold projects.
  • Do work on data you already have: Direct the tool to ‘work’ on projects, directed by natural language.
  • Complex refactoring & debugging: Efficiently reorganise code, resolve warnings, track down bugs.
  • UI E2E automation: Test webflows, troubleshoot UI bugs with screenshot aids.
  • Design system planning: Architect Mode helps you understand and structure large unfamiliar codebases. It can even do translations between languages and make changes.
  • Custom workflows: Use MCP to automate tasks like CI/CD, infrastructure ops, or ticket triage.

How to get started

  1. Install Roo Code
    • VS Code Marketplace: Search for “Roo Code” / “Roo Cline” and click Install. Or use:
      • code --install-extension RooVeterinaryInc.roo-cline
  2. Open and configure
    • Open the Roo sidebar or use Command Palette: Roo Code: Open
    • Connect your preferred AI provider and paste in your API key.
  3. Start chatting
    • Use natural-language prompts.
    • Roo will edit files, run commands, or open the browser—diffs visible before apply.
  4. Customise workflow
    • Add .clinerules to define environment, roles, or project-specific instructions.
    • Explore MCP to integrate external APIs or tools.
    • Define auto-approval rules and draft custom AI personas for consistency.

Why would you use this tool?

  • Full-stack AI teammate: Handles code, terminal, browser—all in VS Code.
  • Tailored to developers: Multi-modes, context pointers, MCP, local control.
  • Model flexibility: Swap AI backends anytime without breaking workflows.
  • Safe and transparent: Diff previews, auto-approve rules, context logs, usage stats.
  • Open & community-first: Open-source (Apache-2.0), GitHub, Discord & Reddit communities.

This is also an awesome way to just work with files on your local system using native APIs like Google Gemini models - and even Chinese models cheaply. I prefer Claude Code as it does the same thing, but this is a much more affordable way to go.

Conclusion

Roo code transforms VS Code into an AI powered coding assistant that combines natural language understanding with file manipulation, terminal control, and browser automation. This versatile tool reduces technical barriers for complex development tasks while offering model flexibility, customizable workflows, and affordability compared to premium alternatives.