Join the Bali AI retreat
wright mode
AI TOOLS·16 NOV 2026

Grok Bot vs Codex: which one should a business owner use?

Grok Bot for jobs handed off and repeated on a schedule. Codex for one-off research and files you'll revise. How cost and decks compare, checked 16/11/2026.

By Brooke Wright · 6 min read

Grok Bot vs Codex: which one should a business owner use?

Use Grok Bot when you want a bot on its own cloud computer to repeat a job on a schedule while you're away. Use Codex when you want ChatGPT to work through a one-off job with you and hand back a file you can keep revising. On price, Codex comes with ChatGPT Plus at US$20 a month, and Grok Bot starts at US$30 a month on The X Subscription's Premium plan. Neither charges per job. Both ration you with usage limits, so the cost of a job is how much of your allowance it eats.

I use both. My desk work happens in Claude Code and Codex, and Grok Bot gets the jobs I want handled from my phone or while my laptop's shut. Below: what the same research job costs, what recurring work costs, and how each handles an editable presentation, all checked against the official pages on 16/11/2026. Neither company publishes a cost per job, so the test near the end gets you your own number.

If you want all the agents side by side, start with my ChatGPT Dots vs Grok Bot vs Meta Muse vs Hermes comparison. This one is just these two.

What's the actual difference between Grok Bot and Codex?

Grok Bot is a team of bots that share one cloud computer. xAI describes it as a persistent cloud computer with a browser, filesystem and terminal. Closing the app, laptop or phone doesn't stop a routine.

Codex lives inside ChatGPT. OpenAI lists it in the desktop app, the mobile app, on the web, in a command line tool, an editor extension and Codex Cloud, with browser, computer use, files and web search alongside. One wrinkle: OpenAI also sells ChatGPT Work, "a way to delegate real work to ChatGPT", and says people who used Codex for non-coding work can stay in Codex or switch to Work. Its page doesn't list which plans include Work, so check your account.

How should I compare the cost of researching the same list of target accounts?

Run the same list through both and compare how much of your allowance each one used. Sticker prices won't tell you, because the two meter usage differently.

Codex's pricing page measures Plus in local messages per five-hour period, and the figure changes by model. It adds that weekly limits may also apply, and that cloud tasks "may use more of your allowance than local messages".

Grok Bot uses one weekly pool. Under The X Subscription, Grok Bot usage draws from that seat's shared weekly pool, which also covers Grok chat, and usage rates vary by activity. In my own setup, bots clicking around websites in their own browser eat usage, and so did a group chat where my bots kept messaging each other.

Account research is mostly browsing and reading, so expect it to be one of the heavier jobs on either plan. Where there's a connected app or a source you can hand over directly, use that instead of letting the agent wander the web.

How do costs differ for recurring work and occasional work?

Occasional work suits the cheaper plan with enough allowance. Recurring work is where frequency decides the bill.

For occasional jobs, Codex on Plus (US$20) is the lower entry price. When you hit the limit, Plus and Pro users can buy more credits. Grok Bot's Premium is US$30, and past the included pool, further usage is Additional Usage, paid from prepaid top-up credit. Auto top-up is off until you switch it on, and you can set a monthly spending cap.

For recurring jobs, Grok Bot is built around routines. A bot can own up to 50 routines, at least five minutes apart, and "each run spends usage". A weekly report is about 52 runs a year. A routine that fires every hour is a very different bill. My rule: run a routine as often as you need, not as often as you can.

Codex can repeat work too. OpenAI says scheduled tasks can repeat, monitor or refresh work over time, such as a daily agenda update.

My pick: weekly reporting you want delivered without thinking about it goes to Grok Bot. A research job you run now and then, then edit by hand, goes to Codex.

How does each one create and revise an editable presentation?

Codex has the better documented editing loop. OpenAI says you can create, preview and refine documents, presentations, spreadsheets and PDFs in ChatGPT. You can point at a specific part of the preview with annotations and ask for a focused change. Its tips: name the exact slide, say what must stay the same, and ask for the new file name before you download the next version. The Codex command line can create and edit files but has no visual preview.

Grok Bot can produce decks too. xAI's own example of an output is a slide deck with speaker notes, and its Google Drive connector can read your past decks and draft the next one in the same format, saving it to Drive for your edits. To change a file, tell the bot to revise the existing one rather than making copies. One catch: saving to Google Drive always creates a new file, even when the name matches. Expect a pile of versions, so name each round.

xAI's page doesn't say which file format a deck comes out as, so ask the bot for the format you need and open the result before you rely on it.

Try this in 15 minutes

  1. Pick a small list: 10 target accounts, or 10 competitors.
  2. Write one prompt: "For each, find what they sell, who they sell to and one recent change. Give me a table with a source link per row. Don't contact anyone."
  3. Note your usage meter in each tool before you start.
  4. Run it in Codex, then in Grok Bot. Note the time.
  5. Check the meter again. Spot-check three rows against their sources.
  6. Ask each to turn the table into a five-slide deck, then ask for one change to slide three only. See which edit is easier to make.

Quick answers

Do both keep working when my laptop is shut?

Grok Bot does. Codex depends on the job. OpenAI says computer use needs your Windows device kept on and connected, and on a Mac locked use is a separate setting you switch on.

Can I choose the AI model?

Not in Grok Bot. xAI says it picks the model for each task, and there is no model picker. Codex's pricing page lists a separate allowance for each model.

Where does Claude fit?

Same job as Codex, built by Anthropic. I compare the two companies in Claude vs ChatGPT for business. If your job is building a site rather than research, read how I built one with Codex.

Want a second pair of eyes on your test?

I run a paid AI membership, so I have an obvious interest here. Comparing agents is quicker when you can run your own job live and ask someone why the output looks off. Inside the membership, we build with agents together on the monthly live Build with Brooke, and the fortnightly Co-Work + Q&A and the private community are there when a setup breaks. Join the Wright Mode membership. If you're starting with Grok Bot, the setup guide is the order I'd do it in.

keep going

Build the AI system your business needs next.

Inside Wright Mode, use my templates, join monthly Build with Brooke sessions and bring your questions to fortnightly co-working and Q&A. Build alongside women putting AI to work in their businesses.