An agent
that grows
with you.
We just host it.
Your own dedicated AI agent on Telegram, ready in 90 seconds. From ₹499 / $6 a month. Bring your own LLM key — pay for the agent, not the model.
Closed beta. We're onboarding users one batch at a time.
- · #142 — session validation moved to middleware
- · #147 — JWT → opaque tokens
- · #151 — MFA gate behind feature flag
- 1. Acme contract review (Sarah, due fri)
- 2. Q2 budget thread (CEO, 8 replies)
- 3. Payments incident postmortem (eng)
Your agent in Telegram. Real session, real cron, your data.
You already know.
Self-hosting Hermes is doable. It's also a small unpaid second job. Here's the side-by-side.
- ✗Wrestle a Python venv on a VPS
- ✗Open ports, set up TLS, hope nothing crashes
- ✗Babysit through OS upgrades + skill updates
- ✗Backups you forgot to schedule
- ✗Memory + sessions wiped on every redeploy
- ✗3am pages no one's paying you for
- ✓Sign up. Paste a Telegram + LLM key. Done.
- ✓TLS, networking, secrets — handled
- ✓Auto-updates, auto-restarts, scaledown built in
- ✓Daily Convex-backed snapshots, one-click restore
- ✓Persistent volume — memory survives forever
- ✓We get paged. You don't.
Not your job. Ours.
From signup to a working agent in five minutes.
No infrastructure to think about. The agent is yours, the data is yours, the bill from your model provider goes straight to you.
Sign up
Create an account. We provision your dedicated agent and get it running.
Pick a plan
Mini to Max. Resize anytime — your agent moves with you, same memory, same chat history.
Connect Telegram + LLM key
Paste a bot token from @BotFather and your Anthropic, OpenAI, or OpenRouter key. Done.
What people actually do with it.
Concrete scenarios people set up the first week. Cron-driven, voice-driven, conversational — your agent does whichever feels natural.
Triage Gmail at 8am
Every weekday morning, Hermes scans unread mail, flags anything urgent, and drafts replies for the top three. You wake up with a clean inbox queued.
Voice memo to action items
Forward a voice note from your phone. Hermes transcribes, extracts action items, and files them in your todo list. No more notes lost in the wild.
Weekly arxiv digest
Every Monday, Hermes pulls the top five AI papers from the last week, summarizes each in a paragraph, and posts a thread. Optional: only papers from labs you follow.
Pre-meeting briefs
Fifteen minutes before each meeting, Hermes drops a 2-line context: who's in it, what was last discussed, what's open. Connect Google Calendar once and forget it.
Engineering memory
"What did I change about auth last sprint?" Hermes reads your git history, finds the PRs, summarizes the decisions. Works for any repo it has read access to.
Personal CRM
"Remember that Sarah from Acme is a designer who hates Slack." Hermes carries it forward to every future conversation about Sarah. No app, no schema — just facts.
Plugs into the things you already use.
Hermes ships with native skills for chat platforms, model providers, productivity apps, and the web. Bring your own API keys; we just keep them encrypted.
- ·Telegram
- ·Discord
- ·Slack
- ·Signal
- ·Anthropic
- ·OpenAI
- ·OpenRouter
- ·Ollama Cloud
- ·NVIDIA NIM
- ·Kilocode
- ·Gmail
- ·Google Calendar
- ·Notion
- ·Linear
- ·GitHub
- ·Todoist
- ·Browser
- ·Search
- ·Vision
- ·Codebase
- ·Terminal
- ·Skills
Plus anything in Hermes' plugin ecosystem — install with /skill install <name> straight from the chat. Full list at hermes-agent.nousresearch.com.
An agent that earns its keep.
Everything you'd build for yourself if you had a weekend, ready to go.
Lives where you do
Telegram-first today; Discord, Slack, WhatsApp, Signal coming next. Pick up a conversation on whichever app is in your pocket.
Grows the longer it runs
Persistent memory and skills survive restarts and resizes. Your agent actually remembers you.
Scheduled automations
Natural-language cron. "Every weekday 9am, summarize unread email and post the top 3." Runs whether your laptop is open or not.
Delegates & parallelizes
Sub-agents handle their own conversations and tasks. Long jobs run in the background; you get pinged when they're done.
Hardened sandboxing
Tools run in isolated containers, never on your machine. The agent can be busy without breaking anything.
Web & browser control
Search, page extraction, browser automation, vision, image gen. Connect any other tools you already use.
One agent. Four sizes.
Pay for the container. Pay your model provider directly. Move up or down at any time.
Mini
- · 1 vCPU
- · 1 GB RAM
- · 10 GB
- · 200 msg/day
- · 500 tool runs/day
Lite
- · 2 vCPU
- · 4 GB RAM
- · 40 GB
- · 1,000 msg/day
- · 2,500 tool runs/day
Pro
POPULAR- · 4 vCPU
- · 8 GB RAM
- · 80 GB
- · 5,000 msg/day
- · 10,000 tool runs/day
Max
- · 8 vCPU
- · 16 GB RAM
- · 160 GB
- · Unlimited msg/day
- · Unlimited tool runs/day
Beyond the four tiers.
Dedicated host, multi-user workspaces, custom skills, signed SLA, invoicing in INR or USD, SSO, audit logs, and a private Slack channel with our team. Pricing depends on scale.
- · Dedicated infra
- · Custom skills
- · SSO + audit logs
- · Signed SLA
- · Procurement docs
Things people ask before signing up.
An always-on AI agent dedicated to you. It lives on Telegram, learns your preferences, runs scheduled tasks, and can use the web. You bring your own model API key — we handle the rest.