If you are trying to run Codex as an actual employee — autonomous, trusted, capable of executing goals that span hours or days — the plan you are on determines whether that is even possible. This is the part most business owners skip entirely. They sign up for whatever ChatGPT plan they already had, open Codex, hit a wall within a few days, and decide the tool does not work.
The tool works. The plan does not.
I have spent 6 to 9 months running Codex across my own business and my clients' businesses. The single fastest predictor of whether someone walks away frustrated in week one is the plan they are on. This post covers how to choose correctly before you build anything.
Watch me explain this live — this section is covered in the setup sequence before anything else is touched.
The Three Tiers and What They Actually Mean for Agentic Use
There are three relevant pricing tiers. What separates them is not features — it is usage headroom. And usage headroom is everything when you are asking Codex to behave like an employee rather than a chatbot.
| Plan | Monthly Cost | Recommended For | Agentic Viability |
|---|---|---|---|
| Basic | $20/month | Exploration, occasional prompting | Not viable for serious agentic use |
| Standard | $100/month | Business owners running Codex operationally | Recommended entry point |
| Heavy Use | $200/month | High-volume, sustained multi-project operations | For users consistently hitting Standard limits |
The $20 plan is where most business owners start because it is where they started with ChatGPT. It made sense for conversational use. For agentic use, it is not a starting point — it is a ceiling you will hit in days.
Why the $20 Plan Fails for Agentic Operation
Codex operates under two usage caps: a rolling 5-hour limit and a weekly limit tied to your monthly plan. On the $20 plan, these caps activate quickly and frequently.
Here is what actually happens. You set Codex a goal — something meaningful, like reorganizing your file system or analyzing 12 months of email data. The task runs. You come back and find Codex has stopped. It tells you the limit has reset at 9 PM. You wait. You resume. You hit another cap before the task finishes.
This is not a bug. It is the plan working as designed. The $20 plan was built for conversational prompting, not for agentic goals that can run 24 to 36 hours or longer.
The experience leads business owners to one of two conclusions: either Codex is unreliable, or Codex is not ready for real use. Neither is true. The plan is the constraint.
Codex can run goals that sustain autonomous operation for extended periods. That is the value. The $20 plan makes that value inaccessible.
The $100 Plan: The Real Entry Point
The $100/month plan is the recommended entry point for business owners who want to run Codex as an operational system. At this tier, the usage headroom is substantial enough that most business owners do not hit their monthly limit in the course of normal use.
That framing matters. If you are running Codex the way it is designed to run — setting goals, letting it execute, reviewing results, iterating — the $100 plan gives you enough runway to work that way for an entire month without rationing usage.
The way I think about this: $100/month is not a software subscription. It is the replacement cost for a class of employee work that would otherwise cost $50,000 or more annually in salary. That reframe changes the math immediately.
If you are paying $20/month and getting frustrated, you are not underpaying for a bad tool. You are underpaying for a capable tool and getting a constrained version of it. The upgrade is not optional if agentic operation is the goal.
How to Know If You Need the $200 Plan
The $200/month plan exists for heavier users. Even at this tier, reaching the 5-hour usage limit requires extreme, sustained activity. Most business owners who start at $100 stay at $100 for a long time.
The signal that you need the $200 plan is straightforward: you are consistently hitting your weekly limit before the month ends, and the work you are asking Codex to do is non-negotiable in volume. If you are running multiple concurrent long-horizon goals across multiple projects, that is a $200 use case.
If you are occasionally hitting the 5-hour cap, that is not a $200 problem — that is a workflow design problem. More on that below.
The Speed Setting Is a Billing Decision, Not a Performance Preference
This is the one setting that works against business owners who do not understand what it controls. Inside Codex settings, there is a speed slider. The natural assumption is that higher speed means better results.
What it actually means: faster credit consumption. Increasing speed directly increases how fast you burn through your usage allowance. It is not a performance upgrade — it is a cost lever. Running Codex at fast speed on the $100 plan is the equivalent of voluntarily shrinking your monthly headroom.
Leave it at Standard. Treat the speed setting the way you treat data roaming on an international trip: default to off unless there is a specific reason the urgency justifies the cost.
For the complete framework on how this fits into the full setup sequence, read the full guide.
Common Mistakes at This Step
Mistake 1: Starting on the $20 plan as a "test" before committing.
The problem is that you cannot actually test agentic use on the $20 plan. You hit the cap before you get to see what Codex can do. The test is broken by design. If you are going to evaluate whether Codex works as an operational system, you need to evaluate it on a plan where it can actually operate. Anything below $100 is evaluating a constrained version of the tool.
Mistake 2: Upgrading to $200 when the real problem is the speed setting.
Before increasing your plan tier, check what speed you are running at. Business owners who burn through their $100 allowance faster than expected and immediately upgrade to $200 often find the problem was running at fast speed, not a legitimate volume issue. Check settings first.
Mistake 3: Treating usage caps as tool failures.
When Codex stops and tells you the limit resets at 9 PM, it is working correctly. The cap is a plan feature, not a malfunction. The response is not to abandon the task — it is to upgrade the plan or restructure the goal into segments that complete within a window.
Before You Touch Anything Else
Plan selection should happen before you configure a single setting, connect a single plugin, or run a single onboarding skill. The reason is simple: the plan determines whether the rest of the setup is useful.
If you run the full onboarding sequence — sandbox, permissions, file organization, home base — on the $20 plan, you will hit your usage limit partway through and stop. Then you upgrade, come back, and have to resume a setup process you did not finish. That is a waste of time.
Make the plan decision first. Upgrade to $100 before you start. Then proceed with everything else.
The next step in the sequence is setting up the sandbox — the local environment where Codex operates on your computer. Learn how to set up your Codex sandbox correctly so you are not redoing foundational steps later.
One Principle to Take Away
Every dollar below $100/month buys you access to the tool, not access to its value. The value is agentic operation — autonomous, sustained, trusted execution. That requires headroom. Headroom costs $100/month. Everything below that is a demonstration of what the tool could do if you paid for the version that actually works.
Decide once. Upgrade before you start. Then build.
— Shanee
*
Use the prompts behind this system
The Growth Academy Skills Dashboard includes 100+ Codex skills and prompts for SMB owners.
See the Skills Dashboard →