Summary
- • OpenAI developing persistent always-on agent platform inside ChatGPT, codenamed Hermes, currently in beta
- • Agents run 24/7 as named teammates with task scheduling, event triggers, and connector-based messaging integrations
- • Platform supports custom workflows and skills; agents can operate and receive triggers outside the ChatGPT window
- • Placeholder CTO and CPO roles hint at multi-agent orchestration per account; no confirmed release date
Details
OpenAI's Hermes agent platform surfaces as a beta inside ChatGPT's Agents area
The section is positioned at the top of the Agents area, signaling first-class product intent rather than an experimental feature. No official release date has been confirmed.
Agents run continuously with scheduling, event triggers, and messaging integrations
The platform includes task scheduling and connector support, enabling agents to act on cadence or incoming messages rather than waiting for a user prompt. Conversations and triggers can reach agents outside the ChatGPT window.
Users assemble custom workflows by attaching skills, connectors, and messaging surfaces
The building blocks resemble a full agent composition layer — users wire together capabilities rather than configuring a single monolithic assistant.
Placeholder agent roles like CTO and CPO hint at multi-agent orchestration within one account
OpenAI appears to be positioning the platform for users to run function-specific agents in parallel, approaching something resembling a small AI-run organization for individual users or teams.
OpenAI's entry puts direct pressure on Notion's Custom Agents product
Notion has been the most visible player in trigger-based, shared agent teammates with permissions and scheduling. OpenAI bringing equivalent functionality to hundreds of millions of existing ChatGPT users significantly raises the competitive stakes.
Hermes branding appears consistently across the ChatGPT web app, indicating active development
The repeated internal codename surfacing suggests a sustained engineering effort rather than a speculative mock-up, though no confirmed ship date exists.
Product Launch = new product or major feature reveal, New Tech = novel capability or architecture, Strategy = business positioning or product direction, Market Impact = competitive or industry implications, Industry Update = development status or organizational signal
What This Means
If Hermes ships as described, OpenAI would effectively collapse the gap between a chat assistant and a persistent workforce — users could delegate ongoing business functions to named, always-on agents rather than issuing one-off prompts. This signals that OpenAI intends to own the full agent layer, not just the model underneath it, which poses a structural challenge for any company building agent products on top of or alongside ChatGPT. The competitive pressure on dedicated agent platforms like Notion is real: OpenAI's distribution advantage among hundreds of millions of existing users makes it very difficult for standalone platforms to differentiate on infrastructure alone.
Sources
- OpenAI develops platform for always-on Agents on ChatGPTTestingcatalog