Version 1.107 is now available! Read about the new features and fixes from November.
Dismiss this update
You can customize chat in Visual Studio Code to match your coding practices and project requirements. Set up persistent configurations that automatically apply your preferred context, tools, and guidelines to every conversation. This saves time and ensures consistent responses without manually providing the same information in each chat request.
There are six main ways to customize chat in Visual Studio Code. These options work independently or you can combine them for more comprehensive customization.
Custom instructions let you define common guidelines or rules in a Markdown file for tasks like generating code, performing code reviews, or generating commit messages. With custom instruction, you describe how a specific task should be performed. VS Code can automatically apply these instructions or you can choose to include them in specific chat requests.
Use custom instructions to:
Agent Skills enable you to teach Copilot specialized capabilities through folders containing instructions, scripts, and resources. This feature is currently available in VS Code Insiders only. Agent Skills is an open standard that works across multiple AI agents, including VS Code, GitHub Copilot CLI, and GitHub Copilot coding agent. This makes skills portable across different AI tools. Unlike custom instructions that primarily define coding guidelines, skills focus on specialized workflows and capabilities.
Use Agent Skills to:
Prompt files let you define reusable prompts for common and repeatable development tasks in a Markdown file. Prompt files are standalone prompts that you can run directly in chat. You can include task-specific context and guidelines about how the task should be performed. Combine prompt files with custom instructions to ensure consistent execution of complex tasks.
Use prompt files to:
Custom agents are a way to create a specialist assistant for specific roles or tasks, like a database administrator, front-end development, or planning. Within a custom agent Markdown file, you describe its scope and capabilities, which tools it can access, and a preferred language model.
Use custom agents to:
Language models let you choose from different AI models optimized for specific tasks. You can switch between models to get the best performance for code generation, reasoning, or specialized tasks like vision processing. Bring your own API key to access more models or have more control over model hosting.
Use different language models to:
MCP and tools let you connect external services and specialized tools through Model Context Protocol (MCP). This extends chat capabilities beyond code to interact with databases, APIs, and other development tools.
Use MCP and tools to:
Different customization methods work best for different scenarios. The following table lists common use cases and the recommended approach:
| Use Case | Approach |
|---|---|
| Project-wide coding standards | Custom instructions |
| Language or framework-specific rules | Custom instructions with glob patterns |
| Specialized capabilities that work across tools | Agent Skills |
| Reusable development tasks | Prompt files |
| Use chat for planning or research | Custom agents |
| Define specialized workflows | Custom agents |
| Complex reasoning and analysis | Language models |
| Bring your own model | Language models |
| Integrate external services | MCP and tools |
You can implement chat customizations incrementally, starting with the simplest options and gradually adding more complexity as needed.
Start by experimenting with different language models to get better results for different types of work. Use the model picker in chat to switch between models - try faster models for simple tasks and more capable models for complex reasoning. This requires no setup and provides immediate results.
Create custom instructions for consistent results across all your chat interactions. Create a .github/copilot-instructions.md file with your coding standards and preferences. This automatically improves all chat responses without extra effort. Create different instructions files for different parts of your codebase using glob patterns to target specific languages or frameworks.
Once you identify repetitive tasks, create prompt files for common workflows like component generation, code reviews, or documentation tasks. These save time and ensure consistency across your team.
When you need to connect external services or perform specialized operations, add MCP servers and tools to extend chat beyond basic code assistance.
For advanced usage, build custom agents that combine specific tools, instructions, and context for particular roles or project phases.