You can download and install Copilot-CLI, and start using it straight away, without any additional configuration. However, you'll find that you can improve Copilot's responses if you spend a little time providing it with guidelines and context, and giving it access to tools that are relevant to your project. This article introduces the various ways in which you can customize Copilot-CLI.
Custom instructions
You can provide Copilot with instructions for how it should respond. Whenever you ask Copilot a question, or task it to perform some work, a copy of these instructions is added to your prompt. This allows you, for example, to provide details of your project's coding standards, without having to manually tell Copilot about them each time you start a conversation.
For more information, see Adding custom instructions for GitHub Copilot-CLI.
Hooks
Hooks let you run your own shell commands at key points during a Copilot-CLI session. By defining hooks, you can automate specific operations to be triggered when certain events occur: such as the start or end of a session, whenever someone submits a prompt, after the agent completes a task, or when an error occurs.
For example, you could set up a hook to automatically run tests after Copilot makes changes to code files.
For more information, see Using hooks with GitHub Copilot-CLI.
Skills
Skills are folders of instructions, scripts, and resources that Copilot can load to improve its performance on specialized tasks. By adding skills to your project, you can give Copilot extra knowledge or tools for particular workflows, technologies, or domains.
For more information, see Creating agent skills for GitHub Copilot-CLI.
Custom agents
Custom agents let you define specific expertise and behavior for the CLI when it works on a particular type of task. Custom agents are run as subagents—separately to the main agent that responds to a prompt—with their own context window. This allows Copilot to offload work to custom agents without cluttering the main agent's context window, and to use the expertise of a custom agent when it's a good fit for the task at hand.
You can define the toolset available to a custom agent, so that the tools the agent can use are appropriate to its role. For example, a custom agent that works as a reviewer would typically not be permitted to make changes to code files.
For more information, see Creating and using custom agents for GitHub Copilot-CLI.
MCP servers
The Model Context Protocol (MCP) allows you to add external tools and data sources to Copilot-CLI. By adding MCP servers to Copilot-CLI you can add functionality such as the ability to:
- Query databases
- Access issue tracking systems
- Integrate with CI/CD pipelines
- Generate design diagrams
- Search specialist documentation sources
- Book tickets online
- Integrate with a calendar application
For more information, see Informationen zum Model Context Protocol (MCP).
Plugins
Copilot-CLI plugins are distributable packages that provide a simple way to extend the functionality of the CLI.
They bundle multiple customization components together into a single installable unit. You can install plugins directly from a repository, from a plugin marketplace, or from a local path.