VS Code Copilot Agent Tools: Your AI Developer Assistant

VS Code's Copilot Chat includes 14 built-in tools that transform your AI assistant from a simple chatbot into a fully-fledged developer with hands. These tools let agents read code, modify files, run commands, and navigate your entire codebase.

Think of it this way: ChatGPT can only talk about code. Copilot agents can work with code.

How to Access Agent Tools

Here's where you'll find the agent tools in VS Code:

The 14 Built-in Agent Tools

Below are all the agent tools available to you. We're going to deep dive into each one so you know exactly how to leverage your agents and build better workflows:

Built-in agent tools overview

Understanding Agent Tools

Agent tools are essentially actions that your Copilot agent can invoke. When you ask your agent a question, it decides which tools to use to solve the problem.

For example:

"Fix all TypeScript errors" → Agent uses problems + edit tools
"Summarize what changed" → Agent uses changes tool
"Find where this function is called" → Agent uses usages tool

Let's break down each tool and when to use it.

The 14 Built-in Tools Explained

🔄 changes — Get diffs of changed files

Lets the agent read what has changed since your last commit or pending changes.

Useful for:

Generating commit messages
Reviewing modifications
Suggesting fixes only in changed code
Creating code review summaries

Example use case:

"Summarize what changed in this commit."

✏️ edit — Edit files in your workspace

The most powerful tool. Allows the agent to apply file edits directly.

Examples:

Modify code
Add or update comments
Create scaffolding
Update config files
Fix lint errors
Refactor code across multiple files

This is where agents truly shine—they can autonomously improve your codebase.

📦 extensions — Search for VS Code extensions

Agent can search and discover VS Code extensions to recommend or install tools and dependencies.

Examples:

"Which Rust tools should I install?"
"Install Python formatter extension"
"Find the Godot VS Code extension"

🌐 fetch — Fetch content from web pages

Agent can read external documentation, guides, and resources from the web.

Examples:

Fetch API documentation
Look up a GitHub README
Get StackOverflow solutions
Retrieve REST API documentation

Essentially HTTP GET + content scraping—powerful for learning from external sources.

🐙 githubRepo — Search GitHub repositories

Used for cross-repository searching on GitHub. Agent can find relevant code snippets from external repositories.

Examples:

"Search for chess engine implementations in GitHub"
"Find similar Godot enemy scripts"
"Show me an example of ML model training"

This is how agents learn from the broader developer community.

🆕 new — Scaffold a new workspace

Creates a new project based on templates with VS Code-specific configurations.

Examples:

Create React project
Create FastAPI project
Create Godot workspace
Create Docker template

Perfect for quickly bootstrapping new projects with best practices.

🌍 openSimpleBrowser — Preview hosted resources

Opens external pages inside VS Code's Simple Browser.

Useful for:

Opening local documentation
Previewing running web apps
Showing tutorials inline
Testing live websites

⚠️ problems — Check errors in files

Reads diagnostics from VS Code's Problems panel (linting, type errors, warnings).

Examples:

"Fix all TypeScript errors"
"Explain the error in file X"
"Suggest fixes for linting issues"

Agent can see exactly what's broken and suggest fixes.

💻 runCommands — Execute terminal commands

Gives your agent terminal execution capabilities. This is powerful—use wisely.

Examples:

Run "npm install"
Run "git status"
Start dev server
Run unit tests
Build projects

Agent can execute build pipelines, deployment scripts, and development workflows.

📔 runNotebooks — Execute notebook cells

Allows execution in Jupyter notebooks and similar environments.

Examples:

Evaluate data science cells
Run ML training examples
Execute analysis steps
Debug notebook workflows

Great for data science and research workflows.

🔗 runSubagent — Chain multiple agents

Allows delegating tasks to another agent context, enabling sophisticated multi-agent workflows.

Examples:

Delegate security analysis to specialized agent
Pass code generation to architecture-focused agent
Combine multiple specialized agents for complex problems

🏗️ runTasks — Execute VS Code tasks

Runs tasks defined in your tasks.json file.

Examples:

Build project
Lint entire repository
Compile code
Launch Godot build
Run deployment scripts

Perfect for automating project workflows.

🔍 search — Search and read workspace files

One of the most important tools. Lets the agent navigate your entire codebase.

Supports:

Keyword search across files
Reading file content
Understanding project structure
Finding relevant code sections

Without this tool, the agent would be blind to your codebase.

❌ testFailure — Get unit test failure info

Reads context from failed tests, helping the agent understand what broke.

Examples:

Explain test failure
Suggest fixes
Locate problematic lines
Debug test errors

Agent becomes your debugging partner.

✅ todos — Manage TODO items

Searches TODO comments and lets agent manage them across your codebase.

Examples:

Collect all TODOs
Generate TODO backlog
Convert todo notes to tasks
Track technical debt

🎯 usages — Find symbol references

Semantic code navigation for finding where functions, classes, and variables are used.

Examples:

"Where does this function get called?"
"Find definition of AttackComponent"
"List all usages of AIController"

Agent becomes your code navigator.

🔌 vscodeAPI — VS Code API reference

Helps agents build VS Code extensions by referencing official API docs.

Examples:

Reference VS Code API
Build VS Code extension
Scaffold extension structure
Debug extension manifest

Which Tools Matter Most?

| Tool | Why It Matters | |------|---| | search | Agent gains repo awareness | | edit | Agent can modify code autonomously | | runCommands | Agent can execute build/test workflows | | githubRepo | Agent learns from external examples | | fetch | Agent accesses external documentation | | usages | Agent understands code relationships | | runTasks | Agent can automate complex workflows |

The Big Picture: AI With Hands

These tools transform Copilot from a passive chatbot into an active developer assistant:

✅ Reads code ✅ Edits code ✅ Runs code ✅ Navigates projects ✅ Fetches external examples ✅ Performs refactors ✅ Builds projects ✅ Uses APIs ✅ Chains multiple agents

It's literally an AI developer assistant with hands.

Instead of asking ChatGPT "How do I fix this TypeScript error?" and then manually applying the fix, you can ask your Copilot agent "Fix all TypeScript errors" and watch it work.

Practical Example: Fixing Code Issues

Here's a real workflow with agent tools:

You: "Fix all TypeScript errors in this file"
Agent uses problems to read all errors
Agent uses search to understand the codebase
Agent uses edit to apply fixes
Agent uses runCommands to verify: npm run type-check
Result: All errors fixed and validated

Without these tools, the agent would just suggest fixes. With them, it executes them.

Building Custom Agents With Tools

Want to create a specialized agent for your workflow? You can configure agents to use specific tools.

Example agent configuration:

Use search for repo awareness
Use githubRepo for cross-repo knowledge
Use edit for code generation
Use runCommands for build automation
Use fetch for external reference docs

Custom agents transform Copilot from a generic assistant into a specialized tool for your exact workflow.

The Future: AI-Augmented Development

As these tools become more sophisticated, the developer workflow shifts:

Before: Write code → Run tests → Debug → Fix manually With Agents: Describe what you want → Agent implements → Agent tests → Agent fixes issues

Copilot agents represent a fundamental shift in how we think about development tools. They're not just smarter autocomplete—they're collaborative partners that can reason about your codebase and make improvements autonomously.

Key Takeaway

VS Code Copilot agent tools aren't just neat features—they're the foundation of AI-assisted development. Understanding what each tool does helps you leverage agents effectively and build custom agents tailored to your specific needs.

The next generation of developer tools will be those that can not just suggest code, but understand your project and take action to improve it.

Related Articles

Check out these other articles on building AI-powered products:

Beyond ChatGPT Chess: Building the Ultimate Chess Analysis App — Learn how to combine computer vision, Stockfish engines, and GPT to analyze chess positions from real boards with superhuman insight.
Building Dark Souls meets Chess with Godot and LLMs — Discover how we built a roguelite chess game using AI tools for code generation while maintaining quality and avoiding AI slop.

Subscribe to the Blog