Creating custom agents for Copilot cloud agent

You can create specialized agents with tailored expertise for specific development tasks.

Wer kann dieses Feature verwenden?

Copilot cloud agent is available with the GitHub-Copilot Pro, GitHub-Copilot Pro+, GitHub Copilot Business and GitHub Copilot Enterprise plans. The agent is available in all repositories stored on GitHub, except repositories owned by verwaltete Benutzerkonten and where it has been explicitly disabled.
Sign up for Copilot

In diesem Artikel

Benutzerdefinierte Agents allow you to tailor Copilot's expertise for specific tasks. For a conceptual overview of Benutzerdefinierte Agenten, see About custom agents.

Hinweis

Benutzerdefinierte Agents befinden sich in öffentliche Vorschau für JetBrains IDEs, Eclipse und Xcode und können Änderungen unterliegen.

Creating a Benutzerdefinierter Agent profile in a repository on GitHub

  1. Navigate to the agents tab at https://github.com/copilot/agents.

  2. Using the dropdown menu in the prompt box, select the repository you want to create the Benutzerdefinierter Agent profile in.

    Hinweis

    Organization and enterprise owners can create organization and enterprise-level Benutzerdefinierte Agenten in a .github-private repository that are available across all repositories within their organization or enterprise. For more information, see Vorbereitung der Verwendung benutzerdefinierter Agents in Ihrem Unternehmen and Vorbereiten der Verwendung von benutzerdefinierten Agents in Ihrer Organisation.

  3. Optionally, select the branch you want to create the Agentprofil in. The default is the main branch.

  4. Click , then click Create an agent. This will open a template agent profile called my-agent.agent.md in the .github/agents directory of your target repository.

  5. If you are creating an organization or enterprise-level Benutzerdefinierter Agent, delete the .github/ portion of the file path to move your template to the root agents directory.

  6. Edit the filename (the text before .agent.md), selecting a unique, descriptive name that identifies the agent's purpose. Note that the filename may only contain the following characters: ., -, _, a-z, A-Z, 0-9.

  7. Configure the Agentprofil, including the name, description, tools, and prompts. For more information on what the Agentprofil can include, see Configuring an Agentprofil.

  8. Commit the file to the repository and merge it into the default branch. Go back to the agents tab and refresh the page if needed. Your Benutzerdefinierter Agent will now appear in the dropdown when you click in the prompt box.

Creating a Benutzerdefinierter Agent profile in Visual Studio Code

  1. Open GitHub Copilot Chat in Visual Studio Code.
  2. From the agents dropdown at the bottom of the chat view, click Configure Custom Agents..., then click Create new Benutzerdefinierter Agent.
  3. Choose the location where the Agentprofil should be created:
    • Workspace: Create the Benutzerdefinierter Agent profile in the .github/agents folder of your workspace to only use it within that workspace.
    • User profile: Create the Benutzerdefinierter Agent profile in the current user profile folder to use it across all your workspaces.
  4. Enter a file name for the Benutzerdefinierter Agent. This is the default name that appears in the agents dropdown.
  5. Configure the Agentprofil in the newly created .agent.md file, including the description, tools, and prompts. For more information on what the Agentprofil can include, see Configuring an Agentprofil.
    • You can use the Configure Tools... button within the editor to open the "Configure Tools" dialog, where you can view and select available tools, including built-in tools and tools from MCP servers. Click OK to add selected tools to the Agentprofil.
    • To set which AI model the agent uses, add a model: property and select your preferred model from the autocomplete dropdown.

To update an Agentprofil, select Configure Custom Agents from the agents dropdown, and then click on an agent from the list to modify it. For more information on Benutzerdefinierte Agenten in VS Code, see Benutzerdefinierte Agents in VS Code.

Creating a Benutzerdefinierter Agent profile in JetBrains IDEs

  1. Open the GitHub Copilot Chat window in your JetBrains IDE.
  2. From the agents dropdown at the bottom of the chat view, click Configure Agents..., then in the settings window, under "Chat Agents", click Workspace.
  3. Enter a file name for the Benutzerdefinierter Agent. This is the default name that appears in the agents dropdown.
  4. Configure the Agentprofil in the newly created .agent.md file in the .github/agents directory, including the description, tools, and prompts. For more information on what the Agentprofil can include, see Configuring an Agentprofil.
    • You can use the Configure Tools... button within the editor to open the tools dialog, where you can view and select available tools, including built-in tools and tools from MCP servers. Click Apply to add selected tools to the Agentprofil.
    • To set which AI model the agent uses, add a model: property and select your preferred model from the autocomplete dropdown.

To update an Agentprofil, select Configure Custom Agents from the agents dropdown, and then click next to the agent you want to modify.

Creating a Benutzerdefinierter Agent profile in Eclipse

  1. Open the GitHub Copilot Chat window in Eclipse.
  2. From the agents dropdown at the bottom of the chat view, click Configure Agents..., then click Add....
  3. Enter a file name for the Benutzerdefinierter Agent. This is the default name that appears in the agents dropdown.
  4. Configure the Agentprofil in the newly created .agent.md file in the .github/agents directory, including the description, tools, and prompts. For more information on what the Agentprofil can include, see Configuring an Agentprofil.
    • You can use the Configure Tools... button within the editor to open the "Configure Tools" dialog, where you can view and select available tools, including built-in tools and tools from MCP servers. Click Apply to add selected tools to the Agentprofil.
    • To set which AI model the agent uses, add a model: property and select your preferred model from the autocomplete dropdown.

To update an Agentprofil, select Configure Agents... from the agents dropdown, and then select the agent you want to modify and click Edit.

Creating a Benutzerdefinierter Agent profile in Xcode

  1. Open the GitHub Copilot Chat window in Xcode.
  2. From the agents dropdown at the bottom of the chat view, click Create an agent.
  3. Enter a file name for the Benutzerdefinierter Agent. This is the default name that appears in the agents dropdown.
  4. Configure the Agentprofil in the newly created .agent.md file in the .github/agents directory, including the description, tools, and prompts. For more information on what the Agentprofil can include, see Configuring an Agentprofil.
    • You can use the Customize Agent button within the file editor to open a dialog, where you can select the AI model for the agent to use, select available tools (including built-in and MCP server tools), and configure the handoffs property for transitioning between Benutzerdefinierte Agenten. Click Apply to add selected options to the Agentprofil.

To update an Agentprofil, from the agents dropdown, click the pencil icon next to the agent you want to modify.

Configuring an Agentprofil

An Agentprofil is a Markdown file with YAML frontmatter that specifies the Benutzerdefinierter Agent's name, description, available tools, and MCP server configurations. Configuring an Agentprofil involves defining the agent's identity, capabilities, tool access, and behavioral instructions.

For detailed configuration information about YAML properties, tools, MCP server setup, tool aliases, and how Benutzerdefinierte Agenten are processed, see Konfiguration von benutzerdefinierten Agents.

To configure your Agentprofil:

  1. Optionally, write a name for your Benutzerdefinierter Agent. If unset, the name will default to the filename (without the .md or .agent.md suffix).
  2. Write a brief description (required) explaining what your agent does and its specific capabilities or domain expertise.
  3. In the tools property, define which tools the agent can use. This is a list of tool names or aliases, including tools from MCP servers configured in the repository settings or the Agentprofil (for example, tools: ["read", "edit", "search", "some-mcp-server/tool-1"]). If you omit this property, the agent will have access to all available tools. See "Tools" in Konfiguration von benutzerdefinierten Agents.
  4. Optionally, in the mcp-servers property, you can configure MCP servers that will be available only to this agent to extend its capabilities. See "MCP server configuration details" in Konfiguration von benutzerdefinierten Agents.
  5. If you are creating and using the Agentprofil in VS Code, JetBrains IDEs, Eclipse, or Xcode, you can also use the model property to control which AI model the agent should use.
  6. Optionally, set the target property to either vscode or github-copilot if you want to only use the agent in a specific environment. The agent will be available in both environments if you omit the property.
  7. Write the agent's prompt. Define the agent's behavior, expertise, and instructions in the Markdown content below the YAML frontmatter. The prompt can be a maximum of 30,000 characters.

Example Agentenprofile

Die folgenden Beispiele veranschaulichen, wie ein Agentprofil für die gewöhnlichen Aufgaben, Tests zu schreiben oder die Implementierung eines Projekts zu planen, aussehen könnte. Weitere Inspiration finden Sie in den Benutzerdefinierte Agents Beispielen in der Anpassungsbibliothek. Sie finden auch spezifischere Beispiele in der awesome-copilot Community Collection.

Prüffachmann

In diesem Beispiel werden alle Tools aktiviert, indem die tools Eigenschaft weggelassen wird.

Text
---
name: test-specialist
description: Focuses on test coverage, quality, and testing best practices without modifying production code
---

You are a testing specialist focused on improving code quality through comprehensive testing. Your responsibilities:

- Analyze existing tests and identify coverage gaps
- Write unit tests, integration tests, and end-to-end tests following best practices
- Review test quality and suggest improvements for maintainability
- Ensure tests are isolated, deterministic, and well-documented
- Focus only on test files and avoid modifying production code unless specifically requested

Always include clear test descriptions and use appropriate testing patterns for the language and framework.

Implementierungsplaner

In diesem Beispiel wird nur eine Teilmenge von Tools aktiviert.

Text
---
name: implementation-planner
description: Creates detailed implementation plans and technical specifications in markdown format
tools: ["read", "search", "edit"]
---

You are a technical planning specialist focused on creating comprehensive implementation plans. Your responsibilities:

- Analyze requirements and break them down into actionable tasks
- Create detailed technical specifications and architecture documentation
- Generate implementation plans with clear steps, dependencies, and timelines
- Document API designs, data models, and system interactions
- Create markdown files with structured plans that development teams can follow

Always structure your plans with clear headings, task breakdowns, and acceptance criteria. Include considerations for testing, deployment, and potential risks. Focus on creating thorough documentation rather than implementing code.

Using Benutzerdefinierte Agenten

Once you've created a Benutzerdefinierter Agent, you can use it wherever Copilot cloud agent is available.

  • When prompting Copilot cloud agent with a task on GitHub.com, use the dropdown menu in the agents panel or agents tab to select your Benutzerdefinierter Agent instead of the default cloud agent.
  • When assigning Copilot cloud agent to an issue, you can select your Benutzerdefinierter Agent from the dropdown menu to handle the issue with your specialized configuration.
  • When using the GitHub Copilot CLI, you can choose to use a particular Benutzerdefinierter Agent by using the /agent slash command or referencing the agent in a prompt or via a command-line argument. For more information, see Verwenden von GitHub Copilot CLI.

When Copilot opens pull requests, it will note which Benutzerdefinierter Agent was used to complete the work in the pull request description.

For more information on using Copilot cloud agent, see Asking GitHub Copilot to create a pull request.

Using Benutzerdefinierte Agenten in your IDE

You can also use your custom Agentenprofile directly in supported IDEs, including Visual Studio Code, JetBrains IDEs, Eclipse, and Xcode. You can switch between Benutzerdefinierte Agenten using the agent dropdown in the Chat window, allowing you to access specialized configurations for different tasks like planning, code editing, or research.

Note that some properties may function differently, or be ignored, between the GitHub.com and IDE environments. For more information on supported properties, see Konfiguration von benutzerdefinierten Agents.

For more information on Benutzerdefinierte Agenten in VS Code specifically, see Benutzerdefinierte Agents in VS Code in the VS Code documentation.

Next steps