Agents Configuration Repository: A Cross-Device Configuration Solution for Unified Management of Claude Code and Codex

Introduces a unified configuration source for managing Claude Code and Codex configurations across multiple machines, solving the problem of tedious and error-prone manual synchronization, and achieving consistency and portability of development environments. The core value lies in enabling more efficient multi-device configuration synchronization and team sharing through version control and automated scripts.

Modern developers often need to switch between multiple devices such as desktops, laptops, and servers while maintaining a consistent AI-assisted development experience with Claude Code and Codex configurations. Manual configuration synchronization is both tedious and error-prone, and this repository provides an elegant solution.

Includes three core workflows:

1. New device initialization: Clone the repository + run install-all.sh for one-click deployment;
2. New project configuration: Generate project-level configuration files (e.g., CLAUDE.md, project-rules.md, etc.) via new-project-claude.sh;
3. Existing device update: Pull repository updates + re-run the installation script to synchronize configurations. Additionally, modular installers (install-all.sh, claude-config/install.sh, codex-config/install.sh) are provided to support selecting the installation scope on demand.

The repository separates configurations by tool:

• Claude configuration package (claude-config/): Contains global settings, system rules, templates, and initialization scripts;
• Codex configuration package (codex-config/): Contains global settings, rule sets, and authentication templates;
• Documentation resources (docs/): Guides such as CONFIG-BOOTSTRAP.md. Local-only files (not shared): For example, Claude's conversation history (history.jsonl), Codex's authentication，token (auth.json), etc., to avoid sensitive information leakage.

Follows three core principles:

1. Configuration as code: Included in version control to enable auditability, rollback, collaboration, and documentation;
2. Layered configuration architecture: Global layer (cross-device universal), project layer (specific project rules), local layer (device runtime state);
3. Minimize manual steps: Automated deployment via scripts—new device initialization only requires cloning + running the script.

Applicable to multiple scenarios:

• Multi-device developers: Maintain consistent AI experience across different devices;
• Team configuration sharing: New members quickly get a unified environment;
• Configuration experimentation and rollback: Test new configurations via branches and roll back if not satisfied;
• Disaster recovery: Clone the repository on a new device to restore core configurations and workflows.

Expansion directions: Add custom rules, expand project templates (support multiple tech stacks), integrate Git hooks; Limitations to note: Avoid hardcoding sensitive information in configurations (use environment variables), pay attention to platform differences (need to detect OS for adjustments), and focus on tool version compatibility (，record tested versions).

This repository embodies the practice of Infrastructure as Code, solving problems such as multi-device synchronization, team collaboration, and configuration auditability. As AI-assisted tools become more popular, this configuration management model will become a standard in developers' toolchains. It is recommended that developers/teams who take AI-assisted development seriously establish a similar system.