OpenAI’s Codex AI Agent gets new features including Internet access

OpenAI’s Codex AI Agent gets new features including Internet access

OpenAI updates the recently launched Codex AI agent with new features such as internet access and the ability to dictate tasks.

OpenAI launched the Codex AI agent in ChatGPT in May, initially only available to Enterprise subscribers. Now the company announces that Codex AI is also available to ChatGPT Plus users. They will now be able to use the tool as well, although they will encounter rate limits during periods of high demand. Additionally, OpenAI introduces several improvements to the AI agent, such as internet access, the ability to update existing PRs, and dictate tasks.

Improvements

The recently launched Codex AI agent, accessible via the sidebar in the ChatGPT interface, has already undergone numerous fixes and improvements. For instance, the limit for differences between tasks has been increased from 1 MB to 5 MB. The maximum duration for installation scripts has been extended from five to thirteen minutes.

read also

Smarter Coding with the Codex AI Agent in ChatGPT

On iOS, multiple bugs have been fixed. Among others, the duplication of follow-up tasks in the task list has been addressed, and the status display of pull requests has been updated. Live activities have also been re-enabled after a previous issue with notifications. Additionally, the mandatory two-factor authentication has been lifted for users who sign in via SSO or social logins.

Internet access

OpenAI has added new features to Codex within ChatGPT, including the ability to grant internet access for specific tasks. This internet access is disabled by default, so users need to enable it manually.

Internet access for the Codex AI agent. Source: OpenAI

Additionally, Pro and Team users can enable access for specific environments, with detailed control over which domains and HTTP methods Codex has access to. According to OpenAI, internet access will soon be available for Enterprise users as well.

Updating PRs and speech dictation

Furthermore, users now have the ability to update pull requests when executing a follow-up task. Moreover, the “speech dictation” feature is available, allowing users to dictate their task instead of typing.

Updating PRs. Source: OpenAI