OpenAI updates Codex with autonomous macOS control, memory, and new plugins to compete with Anthropic's Claude Code.
The AMW Read
This updates the OpenAI case study in the coding segment by signaling a shift from code-generation to autonomous agentic workflows (macOS control/memory) to defend against Anthropic's Claude Code.
OpenAI updates Codex with autonomous macOS control, memory, and new plugins to compete with Anthropic's Claude Code.
OpenAI has released a significant update to its Codex development system, enhancing its capabilities as an autonomous desktop agent for macOS. The update enables Codex to operate and control desktop applications directly on a user's computer, functioning in the background without interfering with other work. Additional features include image generation via gpt-image-1.5, new integrations with tools like GitLab, Atlassian Rovo, and the Microsoft Suite, native web browsing, and the ability to schedule and resume long-term tasks. A key addition is an opt-in memory feature that allows Codex to learn from past interactions, remembering user preferences and corrections to improve future task execution. The functionality is initially rolling out to macOS users signed into ChatGPT, with EU availability pending.
This update represents a strategic escalation in the competitive AI coding agent market, directly targeting Anthropic's Claude Code, which has gained significant developer traction. By enabling autonomous desktop control and persistent memory, OpenAI is moving Codex beyond a conversational coding assistant toward a proactive, general-purpose software development agent. This shifts the battleground from pure code generation to full-stack workflow automation, integrating with core enterprise development and productivity ecosystems. The focus on macOS first targets a key developer demographic, while the planned expansion signals a broader platform play for enterprise AI integration.
From an expert analyst perspective, this is a necessary and aggressive counter-move by OpenAI to defend its developer tooling segment against a formidable challenger. The autonomous app control feature, while currently limited, points to a future where AI agents are granted broad system-level permissions, raising immediate questions about security, oversight, and sandboxing that enterprises will need to address. The memory feature attempts to solve a key pain point in AI-assisted development—context loss between sessions—which could significantly boost productivity for complex, iterative projects. However, the success of this update will be measured by its reliability in real-world development environments and its adoption versus Claude Code's established workflows, making the coming months a critical period for OpenAI's developer tools strategy.
