Copilot Agent Mode is now available in JetBrains IDEs

Key Concepts:

• Copilot integration in JetBrains IDEs
• Code completion
• Inline chat
• Copilot chat tool window (Ask mode, Edit mode, Agent mode)
• Slash commands
• Project context
• Working sets
• Diff view
• Model Context Protocol (MCP)
• GitHub MCP server
• Autonomous code editing

1. Code Completion:

• Copilot provides code completion suggestions as the user types comments or code.
• The user can accept suggestions by pressing the "Tab" key.
• Example: When starting to write a function, Copilot suggests code completions.

2. Inline Chat:

• Highlighting text triggers the appearance of a Copilot chat icon.
• Clicking the icon opens inline chat, allowing users to ask questions and make specific requests related to the highlighted code.
• Example: Asking Copilot to write a descriptive comment for a selected code block.
• The output can be copied or inserted directly at the cursor.

3. Copilot Chat Tool Window - Ask Mode:

• The Copilot chat tool window in "Ask mode" allows users to ask questions about their codebase.
• Users can add specific files for context, which Copilot uses to generate responses.
• Example: Adding files to provide context for a question about the codebase.
• Slash commands (e.g., /explain) simplify common tasks.
• Users can select their preferred AI model for the task.
• The "project context" option provides the entire codebase to Copilot.

4. Copilot Chat Tool Window - Edit Mode:

• In "Edit mode," Copilot makes direct changes to files.
• Users provide a prompt and define a "working set," which specifies the files Copilot is allowed to modify.
• Example: Adding a prompt to modify code and specifying a working set of relevant files.
• A diff view allows users to review the proposed changes before accepting them.
• Users can iterate on the output with additional prompts.

5. Copilot Chat Tool Window - Agent Mode:

• "Agent mode" enables Copilot to operate more autonomously and determine which files need editing.
• Agent mode supports the Model Context Protocol (MCP).
• Example: Asking Copilot to list open issues in the repository.
• Copilot uses the GitHub MCP server and the "list issues" tool.
• Example: Asking Copilot to implement the first feature in the backlog.
• Copilot outlines each step, including reading files, planning its work, and generating output.
• Copilot requests confirmation before running terminal commands to build and validate the output.
• If errors occur, Copilot automatically continues working to resolve them.
• Users can review changes, iterate with additional prompts, and accept the final output.

6. Model Context Protocol (MCP):

• MCP allows Copilot to interact with external services and tools, such as GitHub, to gather information and perform actions.
• Example: Using the GitHub MCP server to list open issues.

7. Autonomous Code Editing:

• Agent mode allows Copilot to autonomously edit code, including planning, executing, and validating changes.
• Copilot requests user confirmation for critical actions, such as running terminal commands.
• Copilot automatically attempts to resolve errors encountered during the editing process.

Synthesis/Conclusion:

The video demonstrates the capabilities of Copilot within JetBrains IDEs, highlighting its features for code completion, inline chat, and autonomous code editing through different modes (Ask, Edit, and Agent). Agent mode, leveraging the Model Context Protocol (MCP), enables Copilot to interact with external services like GitHub, allowing it to perform more complex tasks such as implementing features from a backlog. The video emphasizes the iterative workflow, where users can review changes, provide feedback, and ultimately accept the final output.