Conductor Plugin Adds Antigravity CLI Support

Key points

• 1

  Conductor becomes a conversational plugin: Conductor has shifted from a Gemini CLI extension to a standalone plugin that supports natural language conversations instead of rigid command sequences. This means developers can discuss feature requirements with Conductor as if chatting with an AI assistant, while still maintaining persistent spec.md and plan.md files. For example, if you're building a new feature, Conductor will dynamically update your context and check off tasks in your plan without requiring manual command inputs. This transition preserves the procedural rigor of spec-driven development (SDD) but makes the workflow more intuitive, allowing teams to focus on architecture rather than command syntax.

• 2

  Antigravity CLI integration: The Conductor plugin now works directly with Antigravity CLI through a specific installation command: agy plugins install https://github.com/gemini-cli-extensions/conductor. This enables developers using Antigravity to leverage Conductor's spec-driven capabilities without switching tools. Unlike before, when Conductor was tied exclusively to Gemini CLI, teams can now start a workflow in one tool and continue it in another—like moving from Antigravity to Claude—without losing project context. This portability ensures that your spec.md and plan.md files remain the single source of truth across any tool, meaning your architecture planning stays consistent even when switching between different AI assistants.

• 3

  Preserved context with flexible workflows: Your project's context, including specs and plans, remains persistent and version-controlled regardless of the tool you use. For instance, if you create a spec.md file with architecture details using Conductor in Antigravity CLI, it will still be accessible when you switch to another tool like Claude. This avoids the common problem of losing context when switching between AI assistants, which was a limitation of the previous Gemini CLI-only setup. The plugin ensures that your team can maintain high task completion rates—specifically, Conductor achieved better success rates on complex TerminalBench tasks compared to non-SDD users—by keeping the project state consistent and actionable across tools.

What changed