built on Hermes Agent + Kimi K2

What should your agent work on?

Drop a task. ChatHermes spawns a private agent that keeps working — even after you close the tab. Memory, skills, scheduling, and Telegram out of the box.

Build me a landing page for…
kimi-k2
no signup until you hit send · powered by Hermes Agent + Kimi K2
live preview
scenario 1/3 · Newsletter while you sleep
chathermes.com / your agent
you
— how it works

It works while you're offline.

Three things that ChatGPT can't do — because it's a chat. ChatHermes is an agent.

Tuesday, 10:47 PMbefore bed
Hermes writing a newsletter at night while you sleep

You leave a half-formed idea. It comes back finished.

Drop a voice-note about a newsletter you've been meaning to write. Then go to sleep.

you"Draft something on why slow AI is undervalued. My voice — terse. Reading list pinned."
hermesGot it. I'll have a draft on your Telegram by 7am.
— 8 hours later, while you slept —
research • 18 sourcesdraft • 740wfact-checkstyle-pass
hermesDraft ready. Pinned 3 quotes from your reading list. Tap to edit, or reply with feedback.
Right nowparallel
Hermes spawning many ideas in parallel

It spawns a swarm. You watch it work.

Big asks split into subagents that run in parallel. Each one is visible — no black box.

you"Build me a landing page for an indie game called Aether. Wire it up, deploy it, send me the URL."
hermesOn it. Spawning 4 agents.
copy • drafting herodesign • 3 variantscode • Next.js scaffolddeploy • waiting
hermesLive at aether-gXk2.chathermes.dev. Want me to hook a domain?
— vs. chatgpt

A chatbot is a calculator. ChatHermes is a coworker.

— presence
ChatGPT sleeping vs ChatHermes always working
— memory
ChatGPT forgets vs ChatHermes remembers
— surface
ChatGPT trapped in a tab vs ChatHermes everywhere
— parallelism
ChatGPT one task vs ChatHermes many in parallel
— what makes it different

Four things that make this different.

01

Zero install.

Open the URL. Chat. The agent is already running on a server we operate. No Docker. No SSH. No env vars.

02

Lifelong memory.

Browse, edit, and curate what your agent knows about you. Memory is reviewable, never opaque.

03

Lives where you live.

One agent, every surface. Connect Telegram in one click — same memory, same skills, same voice.

04

Background work.

"Every Monday, summarize my reading list." It does. Forever. Until you change your mind.

— start now

Stop chatting. Start delegating.

Sign up takes 30 seconds. Tell your agent what you want. Close the tab. Come back to results.

Try it now ↑
— open source

Self-host it for free.
Or skip the setup.

ChatHermes is fully open source under AGPL-3.0. The code that powers chathermes.com is the same code you can clone, fork, and run yourself.

—
Self-host
Open source
Hosted
chathermes.com
Setup time
30–60 min
30 sec
Source code
all of it (AGPL-3.0)
same code, hosted
LLM API keys
bring your own
pre-pooled, included
Hermes Agent native
install yourself (5 GB)
pre-installed, ready
Telegram bot
BYO BotFather token
shared @ChatHermesBot
Email delivery
BYO Resend + DNS verify
verified domain, included
Public preview URL
your-domain.com/p/<slug>
chathermes.com/p/<slug>
Backups
DIY (cron + S3)
automatic + point-in-time
Updates
git pull + restart
automatic
What you bring
your servers, keys, time
your email + a credit card
Privacy
your server, your data
encrypted at rest, SOC2-track
{}
— get the code
github.com/chathermes
Clone, run docker compose up, you're done. AGPL-3.0.
→
⚭
— skip the setup
Try chathermes.com
Magic-link sign-in. No credit card. Free tier with 50 messages / month.
→

The hosted version pays for the LLM tokens, the Hermes Agent runtime, the verified email domain, and the public preview surface — so you don't have to.

ChatHermes

The chat that doesn't end when you close the tab.

built on Hermes Agent·Hermes Agent × Kimi K2·Nous Research × Moonshot AI