OpenAI Codex release notes: EEA gets more app access

Codex just became more relevant for European teams that do not want to build AI workflows around a US-first availability table. OpenAI says in its Codex release notes that more app features are rolling out to the EEA, the UK, and Switzerland. For Swedish and Nordic companies, this is not just a regional footnote in a changelog. It decides whether Codex can be tested in real browser and desktop workflows, with the same control questions that already exist in normal IT routines.

The Codex app is OpenAI's desktop app for parallel coding and workflow threads. A coding agent is an AI system that can read project files, propose changes, and sometimes use tools on your behalf. Computer Use means Codex can see and operate desktop apps with approval. An approval gate is a clear pause where a human approves the next step before the agent continues.

Source: OpenAI Codex changelog: Codex app features are available in the EEA, UK, and Switzerland

OpenAI Codex release notes: what changed on June 16

OpenAI lists four changes for the EEA, the UK, and Switzerland:

• Computer Use is now available on macOS and Windows in these regions. OpenAI describes it as Codex operating desktop apps by seeing, clicking, and typing.
• The Codex Chrome extension is now available for browser tasks that need signed-in Chrome context. It can work across tabs in the background without taking over your browser.
• Memories are available, but off by default in the EEA, the UK, and Switzerland.
• Chronicle is available as an opt-in research preview for ChatGPT Pro subscribers on macOS and can help Codex build memories from recent screen context.

Source: OpenAI Codex changelog, 2026-06-16

Why this matters for Nordic AI and automation teams

This makes Codex more useful in the workflows where automation often gets stuck: internal web tools, customer portals, admin panels, Windows apps, and visual bugs that never show up in a terminal. The Codex Chrome extension matters when the problem only exists inside a signed-in web page. Computer Use fits better when a desktop app or GUI bug needs inspection.

It also makes the control questions more concrete. Before Codex can use an app or signed-in website, the team needs to decide which domains, apps, and data are allowed. In practice, that means scoped API keys, secrets in secret managers or environment variables, least-privilege access, redacted logs, human approvals before sensitive changes, and a simple run log that shows what the agent did.

For Hammer customers, this belongs in Tool Forge: do not just write a prompt, build a small governed workflow that can be rerun, reviewed, and improved.

Human step before the test

Do this yourself before asking Codex to work:

• Open the Codex app in an environment where the new regional availability applies.
• Pick one small task in an allowed app or signed-in web page. Do not use production as the first test if the same flow exists in a test environment.
• If the task needs Chrome, add the Chrome plugin through Plugins and confirm that the extension is connected.
• If the task needs Computer Use, allow only the app or website needed for the task.
• Keep Memories and Chronicle off until your team has actively decided what they may remember.

Source for the Chrome step: Codex Chrome extension documentation. Source for the Computer Use step: Computer Use in the Codex app. Source for memories: Codex Memories and Chronicle.

Short example: use the new Codex feature

After the human setup is done, give Codex a short, bounded task. Here is an example for a signed-in admin view where the Chrome extension is already connected:

Use @Chrome to review the order-status flow that is already open. Compare what you see with docs/order-status.md, list any mismatches with source evidence, and propose the smallest safe patch. Do not change secrets, permissions, billing, or production data. Ask before editing files.

Good output should include:

• which page, app, or tab Codex actually inspected
• clear differences between the documentation and the visible flow
• one small proposed change instead of a broad rebuild
• which steps need human approval before Codex continues

What to watch next

The important part is not turning everything on. Start with a workflow where visual inspection genuinely helps, then keep the agent's access narrow. If the result is useful, the next step can be a repeatable run with a clear checklist, logging, and a named human reviewer.

For Swedish and Nordic organizations, this is a practical signal: Codex is moving closer to real work environments in Europe. That means less waiting for regional rollout, but more discipline in how agent access is designed.