№Managed hosting for personal AI agentsBuilt on open-source Hermes
Host your agent.
Give it a life.
Your agent remembers you, teaches itself new skills, and runs errands while you sleep — on a server that never forgets.
i.
The problem · A brief essay
Your chatbot doesn't know you. It never will.
Every morning you open the same tab, explain who you are, what you're building, how you like your answers — then close the tab and let it forget you by lunch.
The chatbot is a tool you pick up. It has no memory, no schedule, no reach. It waits for you to ask, answers, and goes dark.
An agent should be different. It should live somewhere. Remember you. Do things when you're not looking.
A chatbot is a vending machine. An agent is a colleague who happens to live in a server rack.— the thesis, in one line
ii.
What it is · The answer
A living companion that lives in the cloud, remembers you in plain text, and teaches itself what you need.
Memory
Stored as plain Markdown. Yours to read, edit, export, delete.
Skills
Abstracted from your conversations. Named, reusable, shareable.
Presence
Lives on Telegram first. Discord, WhatsApp, email to follow.
Autonomy
Cron jobs in plain English. Runs while you sleep.
iii.
How it works · Four steps
From nothing to a living agent in four steps.
01Step
Pick a plan.
Monthly tokens, storage, how many agents you want running. Upgrade any time.
~10s
clicks
clicks
02Step
Connect a model.
Use the included free Nemotron 3 Super allocation, or paste an OpenRouter key to run any model — Claude, GPT, Llama, anything.
~20s
one paste
one paste
03Step
Link Telegram.
Paste a bot token from @BotFather. That's your agent's front door — no new app to install.
~45s
with BotFather
with BotFather
04Step
Meet your agent.
It wakes up in our US East datacenter, sends you its first message, and asks what you'd like help with.
~15s
say hi
say hi
iv.
Pricing · Two plans
Simple plans. Real tokens. No surprises.
Starter
$12/mo
A thoughtful companion for daily conversation.
✦ 2M included tokens / month
✦ Free Nemotron 3 Super allocation
✦ Telegram connection
✦Memory & skill-building
✦ One agent
RecommendedStart with Pro →
Pro
$29/mo
Any model via your own key. More room to grow.
✦ 5M included tokens
✦ Bring your own OpenRouter key
✦ Telegram connection
✦ Memory, skills, scheduling
✦ Up to three agents
v.
FAQ · What people ask
Questions worth answering.
An open-source AI agent with a persistent memory, a skill-building loop, and autonomous scheduling. HermesDeploy is managed hosting — we run it for you on Fly.io (Ashburn / US East region today, more soon), with memory that survives across restarts.
Anything on OpenRouter — Claude, GPT-4, Llama, Gemini, DeepSeek, Mistral, Nemotron. Included plans run free Nemotron 3 Super by default. Swap models any time without losing memory.
Memory is stored as plain-text Markdown on a per-user volume. Your API keys live in our managed Postgres, scoped so only your account can read them. We don't sell your data and we don't train models on your conversations. Export or delete everything at any time.
Absolutely — Hermes is open source. HermesDeploy exists because running it yourself means managing uptime, tokens, Telegram webhooks, memory backups, and model fallbacks. If you ever cancel, export your memory and skills as a zip and point your own instance at them. No lock-in — that's the whole point.
Give your agent a life.
Connect Telegram tonight. Wake up tomorrow to an agent that already remembers your name.
Start your 3-day trial →No card required · Cancel any time