Build a Free AI Business Dashboard with ChatGPT Codex
How to build a personal AI dashboard with ChatGPT Codex, combining news, brand mentions, tasks, and newsletters into one control tower app.
What is an AI control tower dashboard?
An AI control tower dashboard is a custom app that pulls together the scattered tools you check every day, industry news, brand mentions, social growth, newsletters, and task lists, into one interface. Instead of hopping between Google Alerts, an RSS reader, a newsletter inbox, and a to-do app, you get a single tab that surfaces what matters and lets you archive or act on it. The version described here was built using ChatGPT’s Codex agent, which writes and iterates on the actual application code based on plain-language prompts.
TL;DR
- Codex acts as the builder, turning natural-language prompts into a working local app rather than requiring you to write code by hand.
- The recommended workflow starts with a rough-draft prompt that lays out every major feature, then follow-up prompts to fix and refine each one individually.
- The example dashboard organizes everything into tabs: industry news, brand mentions, reminders, audience tracking, newsletters, and tasks.
- Getting from an initial skeleton to a fully working app took multiple long Codex sessions, including one stretch of over two and a half hours of iteration.
- The newsletter tab requires Google OAuth setup, connecting a dedicated Gmail inbox so the app can read newsletters and extract news without you opening every email.
- Sorting and filtering options (newest, oldest, watched sites first, confidence-based mention matching) make the news and mentions tabs more useful than a raw RSS feed or basic Google Alert.
- The whole project can be built for free using the ChatGPT app in Codex mode, since it’s a local project rather than a hosted product.
Everyone else built a construction worker.
We built the contractor.
One file at a time.
UI, API, database, deploy.
How do you start building a dashboard like this in Codex?
The build starts inside the ChatGPT app in Codex mode, creating a new local project (in this case named “control center”) linked to a folder on your computer. Codex needs a model capable of substantial code generation, and the higher-effort models available in ChatGPT are worth using for this initial build, even if you switch to a lighter model later for small tweaks.
The first prompt matters more than any other step. Rather than trying to perfect one feature at a time, the approach here was to describe the entire app at once: a business dashboard with sections for industry news, saved reminders, social media audience tracking, a newsletter reader, and a task list, each living in its own tab. The prompt also mentioned that the app should eventually connect to tools like Gmail, Google Calendar, Slack, and Granola, even though those integrations wouldn’t work immediately.
This “lay the groundwork first” method deliberately accepts that the first version will be rough. Codex generated placeholder tabs and filler content in around 15 minutes, with several links and features not yet functional. That’s expected. The goal of the first prompt is a skeleton you can react to, not a finished product.
What actually happened during the build process?
After the initial draft, the process shifted to iterative fixing: pointing out broken links, missing functionality, or features that didn’t behave as expected, then letting Codex work through fixes. This back-and-forth is described as running through several long sessions, with individual Codex work sessions lasting 20 minutes, 23 minutes, 31 minutes, and at one point over two and a half hours, before the dashboard reached a usable state.
One overlooked detail during the build was dark mode. It got added late, then set as the default. It’s a small example of how these iterative builds often miss obvious quality-of-life features until you notice them during testing.
What features does the finished dashboard include?
The final version organizes the app into distinct tabs:
Today. A customizable homepage for a daily brief and task overview. In practice, this tab ends up being the least used part of the dashboard, since most of the daily work happens in the individual module tabs.
Industry news. This tab pulls in articles based on sources configured in settings, such as specific companies to watch (in the example, major AI labs) and keywords like “AI,” “artificial intelligence,” “machine learning,” and “neural network.” Articles can be sorted by relevance, newest first, oldest first, or by watched sites first. Each article can be opened directly, saved to reminders, or archived once reviewed, so the “important now” list only shows things you haven’t dealt with yet.
Mentions. Functions like an upgraded Google Alerts. It searches for mentions of your name, brand, or website, and uses confidence scoring to filter out unrelated results, which matters if you share a name with other public figures. Mentions can also be archived once reviewed, moving them into a history log.
- ✕a coding agent
- ✕no-code
- ✕vibe coding
- ✕a faster Cursor
The one that tells the coding agents what to build.
Reminders. A catch-all save-for-later tab. Items get sent here automatically from the news and mentions tabs, or you can manually paste in a link (a tweet, an article, anything) along with a note explaining why you saved it.
Audience tracking. Configured in settings by entering your profile URLs across platforms like YouTube, X, Instagram, Facebook, LinkedIn, threads, and TikTok. The app tracks follower counts over time, calculates growth against a baseline, and totals your audience size across every connected platform.
Tasks. A straightforward to-do list supporting one-time tasks and recurring daily tasks. Completed tasks move into a history log, and the app tracks the percentage of tasks completed each day.
Newsletters. The most setup-intensive tab. It requires creating a Google OAuth client (through Google’s credentials console), generating a client ID and client secret, and connecting a dedicated Gmail inbox used specifically for newsletter subscriptions. Once connected, the app reads incoming newsletters and extracts the news content, so you don’t have to open and click through each email individually.
Is building your own dashboard with Codex worth it?
For anyone who already juggles multiple free tools (Google Alerts, an RSS reader, a newsletter inbox, a separate task manager, and manual social media tracking), consolidating them into one Codex-built dashboard removes a lot of app-switching. It also costs nothing beyond your existing ChatGPT access, since the app runs locally rather than as a hosted SaaS product.
The tradeoff is time. The build described here wasn’t a five-minute prompt-and-done exercise. It took an initial rough draft, followed by extended iterative sessions correcting broken functionality, plus manual setup work for anything touching real accounts (Gmail OAuth being the clearest example). This approach suits people comfortable with some trial and error and who don’t mind going back and forth with an AI coding agent to get individual features working.
It’s also flexible by design. Not every tab is necessary for every user. Someone without a public brand might skip the mentions tab entirely. Someone who doesn’t run newsletters might skip that integration and focus purely on tasks and news. The prompt-first, fix-iteratively method makes it straightforward to drop or expand modules based on what you actually need.
Frequently Asked Questions
What is ChatGPT Codex used for in this context?
Codex is used as a coding agent inside the ChatGPT app that can generate a working local application from a plain-language description, then continue refining that code based on follow-up instructions and bug reports.
Do you need to know how to code to build this?
No coding experience is required to direct the build, since Codex writes the actual code. However, some patience with iterative debugging (describing what’s broken and asking for fixes) is necessary to get every feature fully working.
Why does the newsletter tab require Google OAuth setup?
Because it needs permission to read emails from a Gmail inbox in order to extract newsletter content automatically. This requires creating an OAuth client through Google’s credentials console and connecting it to a dedicated inbox, ideally one used only for newsletter subscriptions.
How long does it take to build a dashboard like this?
Other agents ship a demo. Remy ships an app.
Real backend. Real database. Real auth. Real plumbing. Remy has it all.
The initial skeleton can come together in around 15 minutes, but reaching a fully functional version with working links, sorting, and integrations took multiple extended Codex sessions, some lasting well over an hour, spread across a longer back-and-forth debugging process.
Can this dashboard connect to tools like Gmail, Calendar, or Slack?
It can be prompted to attempt those integrations from the start, though they typically don’t work immediately and require additional setup and iteration, similar to the newsletter feature, before they function reliably.
