.conductor: An AI-Driven Project Management Framework

November 23, 2025 · View on GitHub

Welcome to Conductor, a structured system for managing software projects with an AI agent like the Gemini CLI.

The core philosophy is to provide the AI with a clear, unambiguous protocol to follow, turning it from a simple coding assistant into a proactive project manager. This system ensures a consistent, high-quality development process from initial setup to daily work.

Conductor works by working with you to create a set of Markdown files that describe, and become, the authoritative plan of record, conceptually similar to a specification, and these files become the guardrails that guide your AI agents (used by Conductor) to build and maintain your project or solution.

How It Works

The Conductor methodology is divided into two main phases:

Phase 1: Guided Setup - A one-time, interactive process where the AI acts as a setup wizard. It scaffolds the project structure, helps you select the appropriate style guides and workflows, and creates the core project management files.
Phase 2: Daily Development - For all subsequent interactions, the AI uses the .conductor/prompt.md file to re-assimilate the project's context and rules, ensuring it always adheres to the established workflow for every task.

Getting Started

Conductor can be launched and used one of two ways, the recommended way is to install it as a registered extension within the Gemini CLI. It can also be called from Gemini manually, via prompting.

Installing and Registering the Extension

Prerequisites

To install as a Gemini extension, first make sure you have the Gemini CLI installed.
Because there is not a published executable, you will also need to clone this Conductor repo locally.

Registering

Once the Gemini CLI is installed and the Conductor repo has been cloned, you can register Conductor as an extension via the following terminal command:

gemini extensions link [path to conductor_repo]/extensions/conductor

Launching Conductor as a Gemini Extension

Conductor can be launched within Gemini using one of two commands:

/conductor:init - Used the first time to initialize your project.
/conductor:resume - Used to open an existing conductor project.

Initializing a Project

To create a new Conductor project, navigate to the root directory of where you want to create your project, if you're using Git, this would typically be your repo directory, and then:

Launch Gemini:
```
gemini
```
Lauch the Conductor project initialization by executing the following command in the Gemini CLI window:
```
/conductor:init
```

Loading an Existing Conductor Project

To load an existing Conductor project, navigate to the directory of the project and then:

Launch Gemini:
```
gemini
```
Lauch the Conductor project initialization by executing the following command in the Gemini CLI window:
```
/conductor:resume
```

Manually Prompting Gemini to use Conductor

Alternatively, you can manually prompt Gemini to use Conductor with a prompt such as the following:

gemini -i "I want to set up this project using the Conductor methodology. Your instructions are in the file located at https://raw.githubusercontent.com/keithballinger/.conductor/refs/heads/main/conductor.md. Download that file with curl, then read it. Then begin the setup process as outlined in that file." --yolo --model "gemini-2.5-pro"

To resume a Conductor project manually, use a prompt such as the following:

gemini -i "Start my development session by following the instructions in .conductor/prompt.md." --yolo

Core Files

conductor.md: The master protocol file used only for the initial project setup.
prompt.md: The entry point for all daily development sessions.
plan.md: The AI-generated, human-approved project plan with phases and tasks.
code_styleguides/: Contains the selected code style guides for the project.
prose_styleguides/: Contains the selected prose style guide for all written content.
workflows/: Contains the selected workflow protocol that the AI must follow for all tasks.