Hermes Agent starter
Try Hermes Agent, the self-improving open-source AI agent from Nous Research, with your own Telegram bot and your own model in a few minutes. It runs shell commands, searches the web, remembers what you tell it and writes itself new skills as it works, and you talk to it from Telegram (Discord and Slack work too). Deploying it adds a persistent volume, so its memory, the skills it learns, conversation history and files survive restarts and redeploys and it can be a long-term assistant. Untick the volume and it runs as a sandbox instead, wiped on every restart: good for experiments and disposable tasks. Your settings live in environment variables, so every restart comes back configured and only the agent's state starts over. Bring a key from OpenRouter, Anthropic, OpenAI, Nous Portal or any OpenAI-compatible API; your provider bills the tokens. Only the chat users you list can talk to it, it asks before running risky commands, and a status page shows what it is doing without ever showing a key. It keeps a connection open to Telegram, so it runs on a plan that never sleeps.
The source
This template is an ordinary public repository. Read it, fork it, or open a pull request against it. Deploying pulls its default branch and builds it the same way velixir builds your own code.
velixirnet/template-hermes-agentOnce it is live
The template is just a starting point. Clone it, edit it, then ship over the top with the CLI:
velixir deploy
Or push your copy to GitHub and connect it on the app's Deploys tab: every push ships a release, and each pull request gets a preview. Both run the same pipeline this template just went through.