if i ask codex to verify on browser the functionality about my project, does it count toward agent request or just chatgpt 5h/weekly plan?
#Codex
low silo
bold tokenBOT
Hi @low silo. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:
The knowledge sources don’t explicitly answer this exact scenario, so I’ll stick to what is documented and flag the gaps.
-
Codex vs. ChatGPT “agent mode” limits
- Codex is a separate product that you access via Codex web, CLI, IDE extension, or SDK, even though it’s included with ChatGPT Plus/Pro/Business/Edu/Enterprise plans.[Using Codex]
- ChatGPT “Agent mode” has its own monthly message limits (e.g., Plus: 40 agent messages/month). These limits apply to “agent requests” inside ChatGPT.[Agent usage]
The docs do not say that Codex tasks (including Codex Cloud tasks) consume these ChatGPT Agent-mode message limits, and they treat Codex as a separate surface.
-
How Codex usage is counted
- Codex usage is measured in “credits” per message, with different rates for Local Tasks, Cloud Tasks, and Code Review.[Codex rate card]
- There is also a separate “5‑hour” style Codex Cloud limit mentioned in community discussions, which is distinct from ChatGPT’s agent message limits.[Forum: Codex Cloud]
The docs don’t spell out the exact mechanics of your “5h/weekly” limit, but they clearly treat Codex Cloud usage as its own quota.
-
Your specific question
You’re asking: If I ask Codex (e.g., via Codex Cloud) to open a browser and verify my project’s functionality, does that count as an “agent request” or only against the Codex 5h/weekly quota?
- The available docs do not explicitly state that Codex Cloud tasks are counted as ChatGPT Agent-mode “agent requests.”
- All the quota and pricing material treats Codex usage (local/cloud/review) as Codex-specific usage, not as ChatGPT Agent-mode usage.[Codex rate card; Using Codex]
- Therefore, the best supported interpretation is:
- Codex tasks (including browser verification in Codex Cloud) consume Codex usage/quota (credits / time like your 5h/weekly), not ChatGPT Agent-mode message limits.
- They may still count toward any Codex-specific limits your plan has, but not toward the “Agent mode” 40/400 messages/month caps.
Because the documentation never directly answers “browser verification” or “5h/weekly” in one place, I can’t guarantee there is no coupling behind the scenes. For an authoritative answer on your specific account/plan, you’d need to contact support via the Help Center.
Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:
low silo
@bold token i'm using codex cli and i asked him to verify my project via browser, hope this explanation helps
bold tokenBOT
Let me check the knowledge base for you again. One moment! :books:
The knowledge sources don’t explicitly say how Codex CLI browser verification is metered, so I can’t give a definitive, account‑level answer—only what’s documented.
From the docs:
- Codex has its own pricing and usage accounting (Local Tasks, Cloud Tasks, Code Review) measured in credits per Codex message.[Codex pricing]
- Codex usage limits “depend on your plan,” and you’re pointed to the Codex pricing/limits page for details; it’s treated as a separate surface from ChatGPT Agent mode.[Using Codex]
- ChatGPT Agent mode has its own monthly message limits (e.g., Plus: 40 messages/month) that apply to “agent requests” started inside ChatGPT, not to Codex.[Agent usage]
Nothing in the sources says that Codex CLI tasks (even when they use a browser) are counted as ChatGPT Agent‑mode messages. All the documented limits for Agent mode are scoped to Agent mode itself, and Codex is documented separately.
So, based on what is written:
- Your Codex CLI request to verify the project via browser should count as Codex usage (and any Codex‑specific 5‑hour/weekly/credit limits tied to your plan),
- and there is no documentation indicating it also consumes ChatGPT Agent‑mode message quota.
Because the docs don’t describe your exact “5h/weekly” counters or Codex CLI browser behavior, if you need a precise ruling for billing/limits on your account, you’ll need to contact OpenAI Support via the Help Center.
Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:
low silo
@bold token please define the differences between those:
Local Tasks, Cloud Tasks, Code Review
bold tokenBOT
Let me check the knowledge base for you again. One moment! :books:
The knowledge sources only give a partial picture, so some details aren’t fully defined. Here’s what is documented:
1. Local Tasks
- Counted as “Local Tasks” in the Codex rate card: 1 message ≈ 5 credits for GPT‑5.2 / GPT‑5.2‑Codex, ≈ 1 credit for GPT‑5.1‑Codex‑mini.[Codex rate card]
- These are Codex actions that run in your local environment via the Codex CLI or IDE extension (editing files, running tests, etc.), where Codex is acting on your machine rather than in a remote sandbox.[Using Codex]
2. Cloud Tasks
- Counted as “Cloud Tasks” in the same rate card: 1 message ≈ 25 credits for GPT‑5.2 / GPT‑5.2‑Codex; not available for GPT‑5.1‑Codex‑mini.[Codex rate card]
- These are tasks you delegate to Codex in the cloud, where Codex runs in an isolated sandbox with your repo and environment, and you later review/merge the results.[Using Codex]
3. Code Review
- Counted separately as “Code Review”: 1 pull request ≈ 25 credits for GPT‑5.2 / GPT‑5.2‑Codex; not available for GPT‑5.1‑Codex‑mini.[Codex rate card]
- This is when Codex reviews a pull request, leaving inline comments in your SCM (e.g., GitHub) using its code‑review capabilities.[Using Codex; Code review with Codex]
In short, based on the docs:
- Local Tasks → Codex acting on your local machine (CLI/IDE).
- Cloud Tasks → Codex running work in a remote sandbox.
- Code Review → Codex reviewing PRs and posting review comments.