GitHub Copilot in JetBrains: Demo of MCP and agent mode

Key Concepts

• Model Context Protocol (MCP): A protocol that enables GitHub Copilot to integrate external tools and context directly into the IDE.
• Agent Mode: A feature within MCP that allows GitHub Copilot to operate more autonomously, taking on tasks like implementing issues.
• MCP Server: A server that provides resources or functionalities that Copilot can leverage.
• MCP Prompts: Predefined commands or actions that can be executed through Copilot by typing a slash (/) followed by the prompt name.
• MCP Elicitation: A structured dialogue flow guided by Copilot to gather information or guide users through processes like configuration or troubleshooting.
• Repository Configuration Elicitation: A specific application of MCP elicitation for configuring a repository.

GitHub Copilot in JetBrains IDEs with MCP Enhancements

This summary details the new capabilities of GitHub Copilot when integrated with JetBrains IDEs, specifically focusing on the enhancements brought by the Model Context Protocol (MCP). These features aim to bring external tools and context directly into the IDE, making Copilot more powerful and autonomous.

Agent Mode

• Functionality: Agent mode, supported by MCP, allows GitHub Copilot to act more autonomously.
• Example: A user can ask Copilot to identify open issues in a repository. Copilot understands to use the GitHub MCP server for this task.
• Process:
  1. Copilot lists the open issues.
  2. The user can then instruct Copilot to "implement an issue."
  3. Copilot outlines the steps it will take, including reading files and planning its work.
  4. Copilot will ask for confirmation before running terminal commands for building and validation.
  5. If errors occur, Copilot will automatically continue working to resolve them.
  6. Users can review changes, iterate on the output with further prompts, and accept them when satisfied.

MCP Enhancements: Sampling and Prompts

• Sampling:
  • Configuration: Users can go to Settings > Model > Context > Protocol to configure which models Copilot is allowed to use at both global and server levels.
  • Application: This allows users to call tools with their preferred model and view the results directly within the chat window.
• MCP Prompts:
  • Access: By typing a slash (/), users can browse prompts from their installed MCP servers.
  • Execution: Users can select a prompt, fill in arguments via pop-up windows, and execute them.
  • Resource Handling: If a prompt requires resources like files or images, they are automatically added as references.

Leveraging MCP Servers for Additional Context

• Adding Context: Users can click "Add Context" and then "MCP Resources" to select from available MCP servers.
• File System Server Integration: For MCP servers that provide resources like the file system server, users can use the "@" mention feature.
• Specific Selection: The "@" mention allows users to select specific configuration files or patterns directly from the file system.

MCP Elicitation

• Concept: MCP elicitation guides users through a structured dialogue flow, enabling Copilot to gather necessary information.
• Example: Repository Configuration Elicitation:
  1. Copilot initiates a structured dialogue to guide the user through repository configuration.
  2. Users respond step-by-step.
  3. Persistence: The conversation history persists even if the IDE is restarted, which is crucial for real-world applications.
• Applications: This feature is highly beneficial for:
  • Configuration wizards.
  • Requirement gathering.
  • Guided troubleshooting, where Copilot asks the appropriate questions instead of the user having to guess what information is needed.

Conclusion

The integration of GitHub Copilot with JetBrains IDEs, powered by the Model Context Protocol (MCP), introduces significant advancements. Features like Agent Mode, MCP Sampling, MCP Prompts, and MCP Elicitation empower Copilot to be more autonomous, context-aware, and interactive. MCP Elicitation, in particular, offers a structured and persistent way for Copilot to guide users through complex processes, making it a powerful tool for configuration, requirement gathering, and troubleshooting. These enhancements collectively contribute to a more efficient and intelligent coding experience.