Short answer: OpenAI’s Agent Mode help page says ChatGPT agent is no longer available and points readers to ChatGPT Work for longer, multi-step tasks and finished deliverables. The page still lists former Agent Mode limits, but they are not an active allowance. For Work, check Settings → Usage for the allowance and reset information shown for your account. Work does not have one fixed number of tasks or messages: usage varies with the plan or workspace seat, model, task, and settings. (help.openai.com)
What limits did ChatGPT Agent Mode have?
The Agent Mode help page lists these former monthly allowances:
- Plus: 40 messages per month
- Pro: 400 messages per month
- Business and Enterprise: 40 messages per month
- Business and Enterprise with flexible pricing: 30 credits per message
These are legacy figures on a page that also says Agent Mode is no longer available. Don’t apply them to ChatGPT Work. The page does not establish detailed counting rules for these figures, such as how follow-ups or scheduled runs were counted.
How to check a current Work limit
Work allowance is not a fixed number of messages or tasks. OpenAI says the amount of work available depends on the plan or workspace seat, model, task, and settings. If Work is available in your account:
- Open Settings → Usage in ChatGPT.
- Check the remaining allowance and reset time displayed for your account.
- If a task stops, check that usage information before concluding that you have exhausted the allowance.
- If your included allowance is exhausted, use the reset information shown for your account. If offered, eligible accounts may also be able to add credits; eligible Plus and Pro accounts may be offered a paid instant reset. Availability depends on the account and, for instant resets, billing country. (
help.openai.com)
Who may have access: Work is available on web and mobile for eligible paid plans. OpenAI lists Work usage with Plus and Pro plans and Business Standard and Premium seats. Desktop availability depends on what is included for your plan and workspace. Enterprise and Edu access is for eligible workspaces with Enterprise Key Management (EKM) enabled; workspace permissions can also affect access. Check OpenAI’s Work and Codex availability guidance. If you use a managed workspace and Work is missing, ask your administrator to check workspace eligibility and access settings. (help.openai.com)
Quota, permission, or action confirmation?
These are different issues:
- Usage allowance: Check Settings → Usage for the allowance and reset information shown for your account. A task count alone is not a universal cap.
- Feature access: If Work is unavailable, check whether your plan or workspace is eligible and whether workspace access is enabled. A permission issue is different from exhausting an allowance.
- Action confirmation: A request to approve an action is not, by itself, evidence that a usage allowance is exhausted.
- API limit or error: API usage and billing are separate from a ChatGPT subscription allowance. Check the API’s error response and your Platform limits; API rate limits vary by model and apply at the organization and project levels. (
developers.openai.com)
Can a production API replace Agent Mode?
An API can support a custom multi-step agent workflow, but it is not a drop-in replacement for the ChatGPT Agent Mode experience. For example, OpenAI documents using the Agents API with computer use to navigate a website, collect information, or interact with an application through its browser interface. That browser runs in an OpenAI-hosted environment. Your application must start and manage the session, handle website access approvals and any sign-in needed, follow the session events, and verify the result. This can fit a workflow you are building around a website or application; it does not mean the API automatically operates in a user’s existing ChatGPT session or reproduces every Agent Mode feature. (help.openai.com)
API usage has its own limits and costs. Rate limits can include requests per minute and tokens per minute, vary by model, and are visible in the developer console. Model requests use the selected model’s API rates; OpenAI tools and hosted sandboxes may have separate applicable rates. These are distinct from a ChatGPT subscription’s Work allowance. The Agents API is currently in beta, so confirm capability, environment, billing, and availability for your intended workload before relying on it in production. (help.openai.com)
Decision rule: If you need work completed inside ChatGPT, first confirm Work is available for your plan or workspace, then check Settings → Usage for your allowance and reset options. Consider an API when you are building a custom workflow and have confirmed that its supported tools and environment fit the job. Budget for separate API rate limits, usage charges, and implementation work; don’t assume feature parity with Agent Mode.
Bottom line
The Agent Mode help page retains former limits of 40 messages/month for Plus and Business/Enterprise, 400 for Pro, and 30 credits per message for flexible-pricing Business/Enterprise, but also says Agent Mode is no longer available. For Work, eligibility depends on plan, seat, workspace settings, and product surface; check Settings → Usage for your actual allowance and reset details. A production API may support a custom multi-step workflow, but its capabilities, limits, billing, and implementation requirements are separate from ChatGPT.
Sources
ChatGPT agent — OpenAI Help Center
ChatGPT Work and Codex — OpenAI Help Center
Managing usage with GPT-6 Astra in Work and Codex — OpenAI Help Center
Using Credits for Flexible Usage in ChatGPT (Personal plans) — OpenAI Help Center
Paid weekly Work and Codex rate limit resets — OpenAI Help Center
Agents API (beta) FAQ — OpenAI Help Center
Troubleshooting API usage and spend limits — OpenAI Help Center