GSoC 2026 - "Multi-IDE Integration Enhancement" project interest and questions

[ThemiraChathumina]

Hi @skeshive,

My name is Themira, and I am very interested in working on the "Multi-IDE Integration Enhancement" project. I have experience in IDE plugin development, specifically using the JetBrains plugin SDK, which makes this project particularly exciting to me.

As I review the requirements and analyze the existing Gemini CLI codebase, I would like to clarify a few architectural points to ensure my proposed approach aligns with the team's vision:

1. JetBrains IDE Plugin Implementation
I noticed that the JetBrains AI Assistant plugin can already connect with the Gemini CLI through the Agent Client Protocol (ACP), and there is also a community-developed companion plugin available. For this project, should we focus on improving this existing ACP integration? Or is the expectation to build a completely new, official plugin from scratch that specifically relies only on the integrated terminal to communicate with the user (preserving the CLI-native experience similar to the current VS Code companion)?

2. Common IDE Context Protocol Specification
The project description mentions establishing a "Common IDE Context Protocol specification" and shared context synchronization. I see there is currently a "Gemini CLI companion plugin: Interface specification" that uses an MCP server to interact with the IDE.

• Should the new JetBrains and Neovim plugins adhere to this existing MCP-based specification?
• Alternatively, is the goal to create a brand-new specification (or adopt the Agent Client Protocol - ACP) to serve as the unified standard across all editors moving forward?

3. Zed Integration Completion
Currently, the Gemini CLI supports standard workflows in Zed (code generation, refactoring, debugging, Q&A, and @-mentions) via an experimental ACP agent. However, certain agent panel features like editing past messages, resuming threads from history, and checkpointing are not yet available as per Zed's documentation. Does the project's goal of "Zed editor full integration completion" specifically refer to implementing these missing features for the ACP agent?

Thank you in advance for your time and guidance. Looking forward to your thoughts!

Thanks,
Themira.

[lesteral]

@ThemiraChathumina et al. - IMO, it would be particularly valuable to address #10955 (Add 'available_commands_update' session notification in ACP protocol)

Thanks for your consideration.

[sakshisemalti]

Hi @skeshive, I've been contributing to gemini-cli's IDE integration, PR #21380 fixes multi-root workspace detection for the VSCode extension. I have JetBrains plugin SDK experience and want to work on Idea #4.
Based on my reading of the codebase and the existing discussions, my approach would be to build the JetBrains companion following the existing VSCode companion spec, using the MCP-based protocol for consistency. I'd start with core functionality, open file, apply diff, get context and then add cursor position awareness and bidirectional navigation.
Would this direction align with what you have in mind for the project?

[aniruddhaadak80]

From my point of view, the strongest version of this project is to define a small shared capability contract before going too deep into editor-specific implementation details. I would treat open file, apply diff, current selection, diagnostics, cursor position, and workspace awareness as the minimum interoperable surface across VS Code, JetBrains, Neovim, and Zed. That would also make the Zed completion question easier to reason about because the remaining work becomes a capability gap instead of a separate integration story. A capability matrix plus thin adapters per editor feels like the safest starting point.