Your own Hermes.
Online in 60 seconds.
A private sandbox. Encrypted secrets vault. Three-layer persistent memory across Telegram, Discord, and Slack. Bring your LLM key. $19/mo, flat.
- Dedicated UNIX sandbox
- Secrets encrypted at rest
- Auto-updated to latest Hermes
- New to Hermes? →
- Telegram
- Discord
- Slack
Hermes, without the DevOps.
Hermes is the fastest-growing open-source agent of 2026. Beautiful framework. Gnarly to run safely. We handle the boring parts so you don't have to.
Isolated sandbox per tenant
A dedicated Linux user with its own filesystem, memory directory, and systemd unit. Your agent never shares state with anyone else.
Hermes' three-layer memory, persisted
Working, episodic, and skill memory survive restarts and span every channel. Your agent remembers what you said on Tuesday, plus the skill it learned doing it.
Messenger adapters, pre-wired
Telegram, Discord, Slack, WhatsApp adapters installed and configured. Drop a bot token in, your agent answers from a DM or a group chat in seconds.
vs. self-hosting it.
You can absolutely run Hermes on a VPS. We've done it. Here's what you'd be doing yourself.
Three steps. No DevOps.
- Step 01
Sign in with Google
No passwords, no email loops.
- Step 02
Subscribe ($19/mo)
We provision your private sandbox in seconds. Cancel anytime.
- Step 03
Paste a bot token
Telegram, Discord or Slack. Your agent is online instantly and starts building memory.
One plan. No surprises.
You bring the LLM key. We host everything else.
- Your own Hermes Agent on a private sandbox
- Telegram, Discord, Slack, WhatsApp: connect any combination
- Persistent memory across every chat, channel and day
- Tool use, file storage, scheduled jobs
- Encrypted secrets, dedicated Linux account
- Cancel anytime via Stripe
Cancel anytime. No credit card needed to sign in and look around.
Questions, before you sign up.
What is Hermes Agent?
Hermes Agent is Nous Research's open-source agent framework — the fastest-growing one of 2026 (95K+ GitHub stars at last count). It pairs a reasoning loop with three-layer memory, native tool use, and six messaging adapters. It's model-agnostic, so it runs on Claude, GPT-5, Gemini, Hermes 4, or anything on OpenRouter. We host it on a private sandbox per tenant so you skip the setup entirely.
Do I need to be technical?
No. Sign in with Google, then paste a bot token — we walk you through getting one in about 30 seconds. No servers, no terminal, no code. If you can use a chat app, you can run an agent.
What can the agent actually do?
It chats, remembers, and acts. Out of the box it can use tools, read and write files in its sandbox, run scheduled jobs, and hold context across days and channels — from drafting documents and tracking tasks to answering questions from everything it has learned about you.
How does the three-layer memory work?
Hermes keeps working memory for the current conversation, episodic memory for things that have happened, and skill memory for procedures it has learned. All three persist to disk in your sandbox, survive restarts, and are shared across every channel — so a fact you mention on Telegram is known on Slack.
Can it learn new skills over time?
Yes. Hermes runs a closed learning loop: when it finishes a task, it can distill the steps into a reusable skill and save it to skill memory. The more you use your agent, the faster and sharper it gets at the things you ask for repeatedly.
Which LLMs work, and who pays for tokens?
Anthropic Claude, OpenAI reasoning models (o-series, GPT-5), Google Gemini, and anything on OpenRouter. You bring your own API key, so model usage is billed directly to you by the provider — never marked up by us. The $19/mo covers hosting only.
Can I connect more than one channel?
Yes. Telegram, Discord, Slack, and WhatsApp can all run at once on the same agent. It's one brain with one shared memory, reachable wherever you are — a private DM or a busy group chat.
Is my data private?
Yes. Every tenant gets a dedicated Linux user with its own filesystem and memory directory — no shared state with anyone. Bot tokens and API keys are encrypted at rest with per-row keys. Your conversations live only inside your sandbox; we don't read them and we never train on your data.
Is the agent always online?
Yes. It runs as a managed service in your sandbox, so it stays online and responsive between messages and keeps building memory in the background. We also keep Hermes auto-updated to the latest release, so you never have to patch it.
Can I cancel?
Anytime — one click in the Stripe customer portal. Your data is yours to export, and the sandbox is destroyed after a short grace period.