OpenAI Developing Always-On Agent Platform for ChatGPT

OpenAI is preparing a new agent area inside ChatGPT with the internal codename Hermes. This section is marked beta and is placed at the very top of the Agents area, appearing closer to a core entry point than an experimental add-on feature.

Users can directly create agents inside ChatGPT and keep them running continuously. The product describes this not as a one-off task tool but as a 24/7 operating teammate.

The components also resemble a platform rather than a single feature.

• Custom workflows can be composed.
• skills and connectors can be attached.
• It can link with messaging surfaces so that conversations and triggers can be reached outside ChatGPT as well.
• task scheduling is included, enabling operation that responds to schedules, events, and incoming messages.

Roles such as CTO or CPO are mentioned as examples, hinting at a direction where multiple agents are grouped by function and operated like a small AI organization. This is the same territory as Notion's Custom Agents, unveiled earlier this year, and could intensify competition by layering an agent tier on top of the skills, connectors, and scheduled tasks that ChatGPT already has.

The Hermes label has been repeatedly confirmed across the web app, but no public release timing has been disclosed yet. If it actually launches in a structured form, ChatGPT is expected to become less like a single assistant and more like a platform where multiple always-on agents move in parallel.