Codex app vs Farero: keep up with the work around Codex

Codex organizes chats into projects (folders).

Farero’s Workspaces show at-a-glance status information, e.g. which repository (by repo icon) the changes are for, whether or not the agents are still working, how recently there’s been activity in them, and details about any pull requests. When you return to Farero you can scan the workspace list and know whether a particular workspace is waiting on review or if there’s an agent that’s still working on something before you open its conversation.

You can group workspaces by repo to find all the changes you’ve been working on across multiple branches that relate to each other, or by status to find e.g. all the work marked for review regardless of what repos it belongs to. Filter down to only unread and sort by recent activity to quickly identify where to pick things back up after a trip to the fridge. Or use the board, with workspaces arranged in columns by status. And when one of the agents finishes a conversation about a PR, the PR will still show up in the workspace.

Finding previous conversations & decisions

You can search previous Codex app conversations.

Farero has a more fleshed-out search experience where you can e.g. limit your search by repo, whether anything happened today, or exclude/include archived workspaces. You can limit by speaker and search just what agents have said or just what you said. Searches can be limited to a workspace or find stuff across workspaces by branch name or even the titles of PRs that were worked on. Results are ordered by relevance or by most recent and show who said the matching bits with the matching sections highlighted. The result takes you straight to the matching message, even if it’s in a chat you’ve since compacted.

Agents can also search across your past chats with any agent (Codex, Claude Code, OpenCode…) and find e.g. project requirements you forgot were in effect, but didn’t commit anywhere. After the agent finds a conversation it can read the surrounding discussion too.

If you’ve spent a long time with an agent iterating on a review or debugging session, Farero’s Compact feature lets you start a new chat with selected messages from that conversation. You can leave out the bulky tool results, and the messages you keep aren’t summarised by another model; they’re copied as written.

In Farero you can also fork a workspace into a new branch in the same repository and bring your uncommitted changes along. You can attach selected parts of the conversation to the new chat too, including if you want to try another AI provider on an alternate implementation.

Working on the app with the agent

Like Codex’s built-in browser, Farero’s browser can sit beside the conversation while work is going on. It has its own pane with tabs, and Codex, Claude Code and OpenCode can all control it: open pages, click around, fill in forms and check the result. When something breaks, the agent can inspect console errors and network requests in that same browser.

You can pick an element on the page and leave a comment like “this button is too close to the edge.” Farero attaches a screenshot and the element details to the chat, so there’s less explaining which button you meant.

Cookie import brings cookies from a chosen browser profile into the tabs you open in Farero, which can save signing in again, though some sites need more than cookies. With agent control enabled for your tabs, the agent can work in those signed-in pages too; separate agent-created pages start clean. You can take control back whenever you want.

On the edge channel, a browser activity indicator also shows which pages agents are using and which chat is controlling them.

Working across mobile & desktop

Both Codex and Farero let you continue work on your phone when an agent needs an answer. In Farero you can even give feedback from your phone and view development server previews. Private remote access uses Tailscale, with the agents and dev server running on your Mac. You’ll need Tailscale set up on the devices involved.

Codex’s documented Mac setup requires power, plus an external display when the lid is closed. Once installed, Farero’s keep-awake helper can keep the Mac awake for a chosen duration (or just while agents are working), even when the lid is closed.

Bonus points for how Farero will fall back to a designated hotspot from your phone (assuming your phone has hotspot set up and available, and your Mac is configured to allow it), when your preferred Wi-Fi network isn’t available, and switch back again later.

Working with several agents

If you like juggling different Codex chats to track all the changes in flight, but wish there was more, a Farero workspace brings the chats, files, diffs and review together in one place, per branch:

• Connect your Codex accounts to Farero (or accounts from other AI providers, even several for each provider). Farero keeps track of which account each chat is using and, optionally, can spread new Auto chats between accounts. Your accounts keep their existing permissions and usage limits.
• Follow along with (and steer or queue follow-up instructions to) multiple simultaneous agent chats.
• Add multiple agents (Codex or otherwise) to one workspace. Different AI models sometimes shine for different tasks. Try out Codex agents for implementation, Claude Code to review, and OpenCode to investigate the repo (for instance). Chats in the same workspace share its files; separate workspaces keep independent changes on separate branches.
• Codex can spawn subagents, send them follow-up instructions and collect their results in the parent conversation. Farero also shows those native subagents, including their histories and questions.
• Agents can also coordinate with other Farero chats you already have open. Ask an implementation agent to read a review conversation and send a follow-up to the reviewer, without copying the discussion between them yourself. This works across Codex, Claude Code and OpenCode.
• Save a reviewer’s model, instructions and permissions as a reusable Role, then delegate reviews to it from the main conversation. Its findings come back to that chat, and the helper’s own conversation stays available when you want the detail. Roles can all use Codex, or use different providers for different assignments.
• Use managed Workflows to put those Roles into phases for a repeated process like investigate/implement/review (beta). Helpers can run in parallel, and you can override the Workflow’s role models and instructions for a particular workspace without changing your usual setup. You can also delegate individual tasks without starting a Workflow.

The small things that make daily work easier

Review the whole branch or an individual commit beside the conversation. If part of a change isn’t worth keeping, discard a hunk, a file or a folder’s changes, with confirmation before anything is removed. A workspace’s GitHub PR is there too, with editable Prompts for things like creating a PR or resolving merge conflicts.

Read and edit files alongside the chat, or switch to rendered Markdown when reading a document is more useful than looking at its source. Images can be copied straight from previews, attachments and chat.

Farero automatically remembers each workspace’s tabs, split view and divider size on your device, so they’re there when you come back. Pin chat/file/diff/review tabs, get around with keyboard shortcuts and the command palette, focus on one conversation in Zen mode, and pick light or dark themes. Farero can also configure new worktree checkouts with copied configuration files and setup scripts.

Try Farero and connect Codex.