Galley

Overview

Galley is a local-first orchestration runtime designed specifically for supervised task execution within AI-assisted repository automation. It allows for seamless integration into any project by running locally and managing work as visible changes within a Git repository, while also ensuring all pertinent evidence is collected before making acceptance decisions.

Key Features

• Claude-first Execution: Initially optimized for Claude Code, with the ability to switch to Codex as an alternative model supervisor.
• Evidence Tracking: Records evidence for review prior to each acceptance decision, enhancing the accountability and traceability of automated tasks.
• Developer-Friendly Setup: Galley can be installed directly into any working repository, maintaining flexibility for developers.

Installation and Usage

To get started, implement the Galley binary into your system’s PATH:

curl -fsSL https://raw.githubusercontent.com/shinpr/galley/main/scripts/install.sh | sh

This command fetches and installs the latest release binary directly from GitHub. Alternatively, developers can utilize Go to install Galley directly:

go install github.com/shinpr/galley/cmd/galley@latest

Once installed, the Galley CLI can be utilized for various tasks, such as setting up repositories, validating workflows, and managing task executions.

Plugin and Skill Integration

Galley integrates seamlessly with Claude Code and Codex through plugins, allowing for enhanced functionality like task authoring and repository inspections. Key commands include:

For Claude Code:

/galley:galley Create a Galley task for this feature request.

For Codex:

$galley Create a valid Galley task and queue it for approval.

Core Concepts

Understanding the core concepts of Galley is essential for effective utilization:

• Task YAML: Defines task goals, acceptance criteria, scope, and execution behavior.
• Quality and Environment Profiles: These provide repo-specific checks and constraints necessary for robust automation.
• Asynchronous Task Handling: Supports unattended tasks that can execute in a managed workspace, providing flexibility and efficiency.

Task Workflow

Galley operates on a structured task lifecycle that moves through various states:

1. Drafting the task YAML
2. Queuing tasks
3. Running tasks in a dedicated worktree
4. Supervisor reviews and final decisions

This workflow ensures a transparent and efficient process from task initiation to execution, offering robust feedback loops and state management.

Conclusion

Galley is a powerful tool for developers seeking effective oversight and automation for AI-assisted tasks in repositories. Its flexibility, traceability features, and integration capabilities empower teams to manage their development workflows more efficiently. Whether using Claude Code or Codex, Galley supports a structured and reliable approach to task execution.