GitHub Copilot for JetBrains 新增 OpenTelemetry 配置优化与模型管理功能

此次更新为 GitHub Copilot for JetBrains 工作流带来更强的控制能力和更清晰的使用体验。现在，你可以在 Claude Agent 流程中连接 MCP 服务器和自定义 Agent，为高级场景调整遥测与 Token 设置，并获得更简洁的聊天和模型选择体验。

现在，你可以为 Agent 工作流配置 OpenTelemetry 导出设置。这使插件行为更容易符合组织对于可观测性的要求。你可以在 Settings > Tools > GitHub Copilot > Chat 中进行配置。

现在，你可以为 BYOK 和自定义端点设置默认 Token 上限，包括 maxInputToken 和 maxOutputToken。你还可以通过模型管理控件禁用或启用所有内置 Copilot 模型。

这些选项使插件行为更容易符合组织在成本控制和模型治理方面的要求。

现在，你可以直接在 Claude Agent 流程中使用 MCP 服务器和自定义 Agent。当你在 IDE 中需要专用工具、自定义指令或团队专属工作流时，这将提供更大的灵活性。

如果你依赖于跨项目共享的 Agent 配置，此次更新可帮助你在适应特定仓库需求的同时，保持工作流程的一致性。

Copilot CLI 会话现已支持分叉，包含 /rubber-duck 命令，并会在 harness 中显示待办事项列表。这些新增功能可帮助你拆分工作、梳理实现思路，并在迭代过程中保持进度可见。

对于企业用户，当其组织尚未配置用户级预算时，我们现在会显示已消耗的 AI 积分数量。有关更多信息，请参阅关于用户级预算的文档。

我们还在持续改善聊天、内联聊天和模型选择的日常使用体验。

• 模型与操作选择器：提升一致性，使控件行为更易预测、使用更方便。
• 自定义流程：提升易用性，减少创建和管理配置细节所需的工作量。
• 会话提示：通过将询问用户的问题渲染为 Markdown，并加入明确的用户注意力通知，提升信息清晰度。
• 内联聊天与模型选择器布局：优化布局表现，带来更简洁的交互体验。
• MCP 诊断：改进诊断能力，帮助你更快理解配置和运行时问题。
• Copilot CLI harness 中的 URL 渲染：改进裸 URL 的显示方式，提升聊天输出的可读性。

此次发布还改进了路径处理和会话记录。Copilot CLI 现在可在 macOS 和 Linux 的工作集及快照中更可靠地保留路径大小写。

我们鼓励你试用最新版 GitHub Copilot 插件，并分享你的反馈。你的意见对于我们持续打磨和改进产品至关重要。

你的反馈将推动产品改进。我们希望通过以下渠道了解你的使用体验：

This update brings more control and clarity to your GitHub Copilot for JetBrains workflows. You can now connect MCP servers and custom agents in Claude agent flows, tune telemetry and token settings for advanced scenarios, and work with a cleaner chat and model-selection experience.

You can now configure OpenTelemetry export settings for agent workflows. This makes it easier to align plugin behavior with your organization’s requirements for observability. You can configure this under Settings > Tools > GitHub Copilot > Chat.

You can now set default token limits, including maxInputToken and maxOutputToken, for BYOK and custom endpoints. You can also disable or enable all built-in Copilot models from model-management controls.

These options make it easier to align plugin behavior with your organization’s requirements for cost control and model governance.

You can now use MCP servers and custom agents directly in Claude agent flows. This gives you more flexibility when you need specialized tools, custom instructions, or team-specific workflows in your IDE.

If you rely on shared agent setups across projects, this update helps you keep your flow consistent while still adapting to repository-specific needs.

Copilot CLI sessions now support forks, include the /rubber-duck command, and show a todo list in the harness. These additions help you break down work, reason through implementation ideas, and keep progress visible while you iterate.

For enterprise users, we now display the number of AI credits consumed when their organization has not configured a user-level budget. For more information, see our docs about user-level budgets.

We are also enhancing day-to-day experience across chat, inline chat, and model selection.

• Model and action picker : Improved consistency so controls are easier to predict and use.
• Customization flows : Improved usability so creating and managing setup details takes less effort.
• Session prompts : Improved clarity by rendering ask-user questions as Markdown and adding explicit user-attention notifications.
• Inline chat and model picker layout : Improved layout behavior for cleaner interactions.
• MCP diagnostics : Improved diagnostics to help you understand configuration and runtime issues faster.
• URL rendering in Copilot CLI harness : Improved bare URL display for better readability in chat output.

This release also improves path handling and session recording. Copilot CLI now preserves path capitalization more reliably in working sets and snapshots on macOS and Linux.

We encourage you to try out the latest version of the GitHub Copilot plugin and share your feedback. Your input is invaluable in helping us refine and improve the product.

Your feedback drives improvements. We’d love to hear about your experience in the following channels: