In a recent LinkedIn post, Charlie Hills 🦩 discusses the capabilities of GPT-6 Astra, detailing a step-by-step process for enabling it to interact with and utilize a user’s Mac.
Charlie Hills 🦩 outlines a practical guide for users looking to integrate Astra with their macOS environment. The process involves several key technical configurations, emphasizing the need for user permissions and specific software setups.
“1. Open Codex and pick Astra (if available). 2. Install Computer Use, enable server and skill. 3. Grant Screen Recording and Accessibility.”
Enabling Astra’s Mac Integration
The core of Charlie Hills 🦩’s insight lies in the specific sequence of actions required to grant Astra the necessary access to perform tasks on a Mac. As Charlie Hills 🦩 notes, this involves not only installing specific tools like ‘Computer Use’ but also granting essential system permissions.
Granting Permissions
Charlie Hills 🦩 highlights the critical nature of granting ‘Screen Recording’ and ‘Accessibility’ permissions. These are fundamental for any application aiming to interact with the user interface and control other applications on macOS. Without these, Astra’s ability to perform actions would be severely limited.
Connecting and Directing Astra
Further elaborating on the integration, Charlie Hills 🦩 explains the subsequent steps, which include connecting Chrome and confirming its management settings. This sets the stage for users to direct Astra to perform specific tasks.
“Mention @Chrome or an app, then say the goal. 6. Approve access to the app or website. 7. Sign in yourself, then check the work.”
According to Charlie Hills 🦩, the interaction is initiated by mentioning a specific application or Chrome, followed by a clear statement of the desired outcome. The user then needs to approve Astra’s access to the relevant application or website, and in some cases, complete the sign-in process themselves to ensure the task is executed correctly.
Defining Task Parameters for AI
A significant portion of Charlie Hills 🦩’s post is dedicated to the importance of clearly defining the scope and expected outcome of any task assigned to Astra. Charlie Hills 🦩 argues that providing explicit constraints and a defined success criterion is crucial for the AI to function effectively.
The Importance of Constraints and Goals
Charlie Hills 🦩 provides a template for how users should frame their requests to ensure clarity and efficiency:
“Use [app or page] to achieve [result]. Use an available plugin for the work it supports, then Computer Use for remaining steps and visual checks. Preserve [constraints], check against [success criteria], and show me the result.”
This structured approach, as outlined by Charlie Hills 🦩, ensures that Astra understands not only what needs to be done but also how it should be done, including any limitations or specific standards for completion. The mention of using plugins and then ‘Computer Use’ for subsequent steps indicates a layered approach to task execution, leveraging different AI capabilities.
Further Resources
Charlie Hills 🦩 also shared a link to a ‘Claude-to-Codex guide,’ suggesting that further resources are available for those interested in exploring these AI integrations more deeply. The post concludes with a call to action, encouraging users to repost the information to help others in their network.
📝 About This Content
This article is based on insights shared by Charlie Hills 🦩 on LinkedIn.
📅 Originally posted on September 8, 2026 | View original post on LinkedIn →