Ai Coding · Tool profile

Codex (OpenAI)

OpenAI coding agent spanning app, editor, terminal and cloud workflows.

Plan at a glance

Plans & sources ↓

Codex is included with ChatGPT Free, subject to usage limits. Codex Cloud is not included on Free or Go.

Limits to check

Cloud environments and connected tools require explicit access. Included usage depends on the account and plan.

Capability sources read 9 October 2026. These are documentation observations; no product test was performed.

Browse & compare tools →

Research scope & testing evidence

How we would evaluate it

Codex is worth evaluating when a task takes more than a code completion: inspect a repository, make a change, run checks and return a reviewable result. Parallel work can help with independent pieces, but it also raises the coordination burden when agents touch shared files. Use separate branches or work areas and define what must be verified before a change is accepted. The useful outcome is a patch you can understand and reproduce, not merely several agents reporting success. Compare it against your existing workflow on one complete bug fix or refactor.

Editorial judgement. This analysis draws from the linked documentation and the workflow described here. It is not a benchmark or a report of using the product.

Fit and limits

Fit and limits

Consider it for

Longer repository tasks and independently separable work with clear acceptance checks.

Check before committing

Cloud environments and connected tools require explicit access. Included usage depends on the account and plan.

What the documentation confirms

What the documentation confirms

  • Official documentation describes parallel agents, cloud tasks and code review.
  • ChatGPT-account access and developer integrations are distinct usage paths.

Source read 9 October 2026. Feature access can depend on plan, region and account.

Composer predictions beta

OpenAI’s 9 October release note describes suggested follow-ups for eligible personal Pro accounts in the desktop app. Predictions support local and SSH threads with GPT-6 Astra or GPT-6.1 Sol. Tab accepts text; it does not send it.

This narrow documentation observation was checked 10 October 2026. Prior pricing and evaluation dates remain unchanged. Read eligibility, settings and usage scope →

Current access and Pro plans

Selected official sources read .

Codex is included with ChatGPT Free, subject to usage limits. Codex Cloud is not included on Free or Go.

PlanBilling basisScope
FreeNo subscription chargeCodex included with usage limits; Cloud excluded
GoSubscription amount not reverifiedCodex included with usage limits; Cloud excluded
Pro 100USD 100/month; monthly billingRevised Work/Codex allowance; numeric total not established
Pro 200USD 200/month; monthly billingExisting lower allowance raised to match higher allowance
Pro 500USD 500/month; monthly billingRevised allowance; Ultrafast available

Cloud requires an eligible paid plan and workspace access. Numeric allowances were not established; no unlimited or permanent-access promise.

Review scope: public documentation only. No purchase or account workflow tested. Plus and Business prices retain their historical date below.

Official access documentation · Pro plans and billing · Source receipt · 29 October allowance update

Historical price observation: 26 October 2026

Historical snapshot. See current access and Pro plan checks from 1 November.

Access and plan costs

Selected official information checked .

PlanBilling basisAllowance
Plus$20/month; monthly billingIncluded usage, not unlimited
ProFrom $100/month; monthly billingHigher included usage
Business$20/user/month equivalent; annual billing2+ seats; $240/user/year calculated
Free / GoLimited trial accessDuration and account eligibility may change

Free and Go have limited trial access, not a promised permanent free Codex allowance. Local and cloud work share usage; API-key billing is separate.

Official source · Codex access and usage

Pro starts at $100/month with monthly billing. Business: $20/user/month equivalent, billed annually for 2+ seats ($240/user/year, calculated). English US page; dollar currency code not explicit. No purchase, invoice or product workflow tested.

Editorial coverage

Editorial coverage

This profile has no applicable recorded five-criterion assessment. It carries no editorial verdict, pick or numerical quality score.

Try one complete task

Give it a small refactor with a behavioural check. Require a diff, the commands run and evidence that behaviour stayed the same.

Create a trial worksheet →

Compare requirements

Edited by Clinton Feyisitan · Evidence policy · Suggest a correction