Hermes Agentvs ChatGPT dots

Hermes Agent works with any OpenAI-compatible API. One Nous Portal login covers hundreds of models, and OpenRouter and other providers work too. Local models run through Ollama, vLLM, llama.cpp or SGLang. You can also use ChatGPT sign-in for Codex models, or sign in with SuperGrok or X Premium+. The /model command switches models mid-session, and conversation history, memory and skills carry over regardless of which provider you use.

Dots, OpenAI's always-on agents in ChatGPT, run on models from OpenAI's GPT-6 family: GPT-6 Astra, and since November 3, 2026, GPT-6.1 Terra. OpenAI says a dot can draw on different models for different parts of its work, and documents no model picker for the main dot. Enterprise model controls and defaults do not apply to it. Tasks a dot delegates to ChatGPT Work or Codex follow those products' own controls.

Hermes Agent is open source under the MIT license, which allows commercial use. You install it where you want it to run: your computer, a VPS or a GPU cluster. Hermes Desktop runs on macOS (Apple Silicon), Windows and Linux. Its commands can run locally, in Docker, over SSH, or in Modal, Daytona or Vercel Sandbox.

Its scheduler takes plain language or cron syntax and can deliver results to your chats. For an agent that must stay up, the docs recommend a server or VPS. Hermes Cloud runs it on a dedicated, always-on instance in the nearest available region. Every paid Nous Portal plan includes one Hermes Cloud agent; larger or extra agents bill daily from credits, and without a plan you need a small credit balance.

Each dot has its own cloud computer, with a Linux system and Chrome browser that OpenAI maintains, so cloud work continues while your devices are off. OpenAI documents no source code or self-hosting option for dots. You can connect your own computers, which must stay online with the ChatGPT app open, and the work is still coordinated in OpenAI's cloud.

One Hermes Agent gateway connects to more than 20 messaging platforms, including Telegram, Discord, Slack, WhatsApp, Signal, SMS, email, Matrix, LINE, Microsoft Teams and iMessage through BlueBubbles. It can talk in Discord voice channels. Several people can share one agent, and allowlists and DM pairing control who can use it. A Hermes Agent mobile app is in beta and coming soon; today you message it through those chat apps, or use Hermes Desktop in your phone's browser with the Webapp.

You can message or call a dot in ChatGPT and message it in Slack, where you can send a direct message or mention it in a channel thread. OpenAI also lists Microsoft Teams, and its Help Center says dots in Teams have limited availability. As of November 9, 2026, the Help Center calls texting a limited beta for Pro users in the US, Canada and Mexico, not supported in Business or Enterprise workspaces.

Hermes Agent keeps a curated memory in two files, MEMORY.md and USER.md, in its home directory, and adds, edits or removes entries as it learns. You can open both files to see what it has saved. Every CLI and messaging session is stored in a local SQLite database with full-text search. When it works out a non-trivial workflow, it saves the method as a skill and improves it with use. You can require human review before skill changes take effect.

A dot receives memories from ChatGPT and creates its own, including from connected apps. You currently cannot view, edit or delete individual dot memories. Disconnecting an app or turning off ChatGPT Memory does not delete what the dot already has; deleting the dot does.

Hermes Agent stores conversations, memory and skills in its home directory on the machine that runs it (~/.hermes on macOS and Linux). Model requests go only to the providers you configure, and a local model needs no network access once it is downloaded. Hermes Agent collects no telemetry by default; usage-metrics sharing is opt-in. Saved passwords are encrypted on that machine, and the model never sees them.

Dots work in OpenAI's cloud, and conversations, tool results and other task context do not stay only on a connected computer.

Hermes Agent runs up to 10 subagents at once by default, each with its own context and terminal session, and you can raise the limit. Profiles run several independent agents on one machine, each with its own memory, skills and keys. In Hermes Desktop, Bot Mode gives each bot its own role, model, memory and skills, and a Group Chat holds 2 to 6 bots. A Kanban board lets named agents share work.

A dot splits work among background agents that run in parallel and report back, and it can delegate parts of a request to ChatGPT Work or Codex. Each member creates one personal dot, and OpenAI says you will be able to add more dots in the future. Specialist dots with their own identity are in focused enterprise pilots.