ChatGPT Agent is OpenAI's built-in agent mode inside ChatGPT that can browse the web, use a virtual computer, and carry out multi-step tasks on your behalf — from researching a topic and compiling results to filling out forms — aimed at everyday ChatGPT users who want the assistant to actually complete tasks, not just answer questions.
Who it's for
ChatGPT Agent is built for individual professionals, researchers, and knowledge workers already using ChatGPT who want it to take real-world actions — browsing sites, comparing options, drafting a document from live web research — rather than developers looking to embed agent behavior into their own product. It's not aimed at teams needing deep, custom integrations into internal business systems; for that, dedicated workflow or enterprise agent platforms are a better fit.
How it works
When given a task, ChatGPT Agent can open a browser inside a secure, sandboxed environment, navigate websites, click and type, read page content, and combine that with reasoning and tool use to complete multi-step work, then report back with results. It operates semi-autonomously: for sensitive actions such as submitting a form with personal data or completing a purchase, it pauses and asks for user confirmation rather than acting unsupervised. It's built into the regular ChatGPT interface rather than requiring a separate app or setup.
Pricing
ChatGPT Agent is available to paid ChatGPT subscribers starting on the Plus plan, priced at $20 per month flat, with no free tier for this specific capability. Usage limits and availability have varied by plan tier since launch, so it's worth checking OpenAI's current ChatGPT pricing page to confirm which plan includes agent access and any usage caps before subscribing specifically for this feature.
Strengths and trade-offs
The strength of ChatGPT Agent is accessibility and integration into a tool many people already use daily — there's no separate signup, API key, or technical setup required, and it combines web browsing with ChatGPT's general reasoning in one place. The trade-offs: it's flat-fee and consumer-oriented rather than usage-metered or enterprise-governed, so it doesn't fit teams needing audit trails, SSO, or fine-grained permissions the way enterprise agent platforms do, and its autonomy is intentionally capped with human-confirmation checkpoints, which is safer but slower than fully autonomous alternatives. For individuals who want an AI assistant that can actually go do something on the web, ChatGPT Agent is one of the most accessible entry points, alongside competitors like Perplexity's Comet and Manus.