A chat window is a tool you open. An agent is something that’s there: you message it from your phone, it remembers what you told it last week, and at 7:30 every morning it sends you a briefing you didn’t have to ask for.
Hermes Agent, Nous Research’s open-source agent, is built for exactly that. The hard part isn’t the agent, it’s keeping it alive.
What an always-on agent does for you
- Answers where you are. Message it on Telegram from anywhere and get an answer from an agent that has your context.
- Remembers. Hermes curates its own memory and searches past sessions, so you stop re-explaining yourself.
- Works on a schedule. “Every weekday at 7:30, summarize my feeds and today’s calendar.” Cron jobs in plain language, delivered to your chat.
- Learns routines. After a multi-step task, Hermes can save it as a skill and reuse it next time.
Setup in four steps
- Run Hermes somewhere that stays on (more below).
- Connect a model. Nous Portal, OpenRouter, OpenAI, or your own endpoint. To survive rate limits, point it at a router that falls back between providers; see one API key for every AI model.
- Create a Telegram bot. Message
@BotFather, run/newbot, and copy the token into Hermes’ Telegram settings. Restrict it to your own user ID so nobody else can talk to your agent. - Give it a first job. Ask it in chat: “Every morning at 7:30, send me the top five posts from these feeds.” Hermes schedules it and delivers there.
The three things that keep it alive
Most personal agents die in week one for boring reasons:
- The process stops. The laptop sleeps, the terminal closes, and scheduled jobs silently don’t run.
- Memory is lost. A container without a persistent volume starts fresh after every restart, and the agent forgets you.
- It can’t be reached. Channels and the dashboard need a stable HTTPS address, not
localhost.
Solve all three and you have an agent. Miss one and you have a demo.
Hermes on a Pod
On Cadesia, Hermes Agent runs in its own Pod: a process that stays up, a persistent disk for memory and skills, and its own HTTPS Address for the dashboard and channels. The AI Agent Starter Solution is Hermes on its own. The AI Agent Router Solution adds 9Router for model fallback at a bundle price.
Still deciding between running your own agent and a hosted assistant? See Hermes Agent vs ChatGPT.
Questions
What can a personal AI agent do on a schedule?
Anything you can describe as a recurring job: a morning briefing of your feeds and calendar, a nightly summary of a channel, a weekly check of a site or repository. Hermes runs these as cron jobs and delivers the result to a connected channel.
Does Hermes Agent remember previous conversations?
Yes. It keeps agent-curated memory, searches past sessions, and builds a model of your preferences over time. That only works if its storage persists across restarts.
Which chat apps can I use to talk to Hermes?
Hermes' gateway supports Telegram, Discord, Slack, WhatsApp, Signal, and the CLI, all from one process. Telegram is the quickest to set up because it only needs a bot token.