In a recent LinkedIn post, Charlie Hills 🦩 offers a strategic approach to integrating AI coding assistants, specifically Claude Code and Codex, for a more robust development workflow. Rather than viewing them as competing tools, Charlie Hills 🦩 advocates for their synergistic use, positioning Codex as the primary engine and Claude Code as a crucial advisory resource.
Charlie Hills 🦩 begins by outlining the setup process, emphasizing the importance of proper configuration for Codex.
“I start in Codex. Claude Code is my backup.”
The initial setup involves downloading and signing into Codex, selecting a project folder, defining the goal, and specifying the necessary files. For Claude Code, the process includes installing its desktop app and navigating to the Code tab within the same project directory. This dual setup allows Codex to handle the core task execution while Claude Code provides a secondary perspective when needed.
Integrating Development Tools
A key aspect of Charlie Hills 🦩’s strategy involves leveraging plugins to extend the capabilities of these AI coding assistants. The process, as detailed by Charlie Hills 🦩, includes searching for and installing desired tools via the Plugins menu, connecting accounts, and initiating new tasks to utilize these integrated tools.
Charlie Hills 🦩 notes the versatility of plugins:
“A plugin can include skills (saved instructions), app connections and MCP tools.”
For more advanced configurations, such as manual MCP setup, Charlie Hills 🦩 directs users to the Settings menu. The post also addresses potential connection issues, advising users to request computer use and grant necessary permissions, such as Screen Recording and Accessibility on Mac systems. Charlie Hills 🦩 shares a personal anecdote about using the Figma plugin and also successfully testing computer use functionality, albeit with a longer execution time.
Refining and Approving AI-Generated Code
The final stage of the workflow, according to Charlie Hills 🦩, focuses on checking and validating the AI’s output. This involves using Claude Code for expert advice, providing constructive feedback to Codex, and rigorously testing the completed work against the initial brief.
Charlie Hills 🦩 stresses the importance of human oversight:
“Test the finished work against your brief. Inspect it yourself before you approve it.”
The author concludes by reinforcing the idea that the ultimate decision-making authority rests with the human developer. To enable direct communication between Codex and Claude Code, Charlie Hills 🦩 also mentions the necessity of installing the Claude Code CLI, with detailed instructions available in a linked migration guide.
This comprehensive guide, shared by Charlie Hills 🦩, provides a practical framework for developers looking to maximize the efficiency and effectiveness of AI coding tools by treating them as complementary rather than interchangeable resources.
📝 About This Content
This article is based on insights shared by Charlie Hills 🦩 on LinkedIn.
📅 Originally posted on September 10, 2026 | View original post on LinkedIn →