Host Hermes, the proven default
AgentRoost's longest-running agent runtime: persistent memory, connector hooks, and scheduling — fully managed on Telegram, live in about two minutes.
What a Hermes session looks like
Built for unattended, long-running work
Remembers across days
Vector stores, conversation history, scratch files — all on persistent NVMe disk. Restarts keep everything.
Hooks already wired
Connector and scheduling hooks come pre-wired; bring your prompts and keys and start.
350+ LLM models
Metered against included credits; switch models any time from the dashboard.
The default for a reason
Our longest-running runtime — pick it if you're unsure; it just works.
Three steps to live
- 1Pick a tier on Pricing — Starter is plenty to begin. Polar handles checkout.
- 2In the wizard, keep Hermes as the framework (it's the default), name your agent, pick a model.
- 3One click connects Telegram via our manager bot — and you're live.
- Starter — $19.99/mo · 1 vCPU · 2 GB
- Plus — $49.99/mo · 2 vCPU · 4 GB
- Pro — $99.99/mo · 4 vCPU · 8 GB
- Monthly LLM credits
- Auto-provisioned Telegram bot
- Persistent NVMe disk
Quick answers
Hermes or OpenClaw?
Both produce the same product — a private 24/7 Telegram assistant. Hermes is the proven default; OpenClaw is the open-source option. Pick Hermes if you're unsure.
What happens to memory on restart?
Nothing is lost — state lives on the workspace's persistent NVMe disk, and supervised restarts reload it automatically.