[Feature Request] Bring Atlas-grade browser features to Codex Desktop

What feature would you like to see?

Please bring more of the browser capabilities previously associated with Atlas into the Codex / ChatGPT desktop app, so the built-in browser can serve as a practical daily workspace rather than only a temporary browser-automation surface.

The goal is to reduce unnecessary context switching between Codex and an external browser while preserving explicit user control, security boundaries, and privacy.

Current pain points

• URLs clicked in Chat frequently open in the system browser, even when the user would prefer to keep the task inside Codex.
• The built-in browser lacks a persistent personal navigation layer such as bookmarks / favorites.
• Users repeatedly jump between Codex and Chrome for documentation, dashboards, GitHub, research pages, and authenticated services.
• Browser state, saved sites, autofill, and extension-based workflows are fragmented between the in-app browser and Chrome.
• The existing Chrome integration is useful for browser control, but it does not yet provide a seamless, browser-like workflow across Codex and Chrome.

Proposed capabilities

1. Configurable link routing

Add settings and context-menu actions for:

• Open Chat links in the Codex built-in browser by default.
• Open links in the system browser by default.
• Ask each time.
• Remember the choice per domain.
• Explicit actions such as Open in Codex Browser and Open in Default Browser.

The setting should apply consistently to links from Chat messages, terminal output, file previews, search results, and other Codex surfaces.

2. Bookmarks / personal favorites

Add a lightweight bookmark system to the built-in browser:

• Bookmark the current page.
• Bookmark folders and a bookmarks bar or sidebar.
• Search bookmarks from the command palette / quick open.
• Optional sync across Codex desktop installations.
• Import from and export to major browsers.
• Allow Codex to suggest or open bookmarks only with clear user intent and permission.

This would be especially valuable for frequently used documentation, project dashboards, GitHub repositories, experiment trackers, internal tools, and research resources.

3. Better continuity between Codex and Chrome

Improve the official Chrome integration so users can intentionally choose between:

• Operating the current Chrome tab / profile through the Codex Chrome extension.
• Opening the same URL in the Codex built-in browser.
• Moving or copying a tab between Chrome and Codex.
• Reusing an authenticated browser session where technically and securely possible.

It would also help to provide clearer connection status, profile selection, tab ownership, permission prompts, and recovery when the extension or native host disconnects.

4. Browser extension support

Consider a controlled extension model for the built-in browser, ideally compatible with selected Chrome extensions or a curated extension API.

Important use cases include:

• Password managers.
• Reference managers.
• Translation tools.
• Developer tools.
• Accessibility extensions.
• Documentation and productivity integrations.

This should use explicit permissions, a visible extension manager, per-site controls, and a safe allowlist / compatibility model rather than silently loading a user's entire Chrome extension environment.

5. Autofill, password managers, and passkeys

Add first-class support for standard browser convenience features:

• Address and form autofill.
• Password-manager integration.
• Passkeys / WebAuthn.
• One-time-code autofill where supported by the OS.
• Clear control over what the agent can and cannot read or submit.

Sensitive fields should remain protected from agent access by default, with explicit confirmation for any action that could expose or submit credentials or personal data.

6. Persistent browser workspace

Bring over more mature browser-workspace features, such as:

• Persistent tabs and tab groups.
• History and recently closed tabs.
• Downloads management.
• Find-in-page and reader mode.
• Profiles or task-specific browser workspaces.
• Restore browser state after restarting Codex.

Why this matters

Codex increasingly combines coding, research, browser automation, file work, and communication in one desktop environment. A more complete built-in browser would make these workflows substantially more coherent:

• Less app switching and fewer interruptions.
• Better continuity between a conversation, repository, terminal, and webpage.
• Easier reuse of trusted project resources.
• More practical browser-assisted coding and research workflows.
• A clearer distinction between user-driven browsing and agent-driven browser automation.

Suggested acceptance criteria

A useful first milestone could be:

1. A configurable default destination for web links.
2. Open in Codex Browser / Open in Default Browser context-menu actions.
3. Persistent bookmarks with folders and search.
4. Reliable Chrome-extension connection status and tab handoff.
5. Secure password-manager / autofill integration with protected sensitive fields.
6. Persistent tabs and session restoration.

Additional note

This request is not asking Codex to silently inherit all Chrome data or permissions. The preferred design is an explicit, permissioned, user-controlled browser workspace that offers Atlas-grade convenience while maintaining strong isolation between user browsing data and agent-accessible browser state.