FlyHermes Explained: What It Is, What It Costs, How To Set It Up And How The Models Behind It Score

FlyHermes is a paid, hosted service that runs a Hermes Agent for you, so you message your own agent on web chat, Telegram, Slack or WhatsApp and never touch a server.

It is operated by HF Data Studio LLC, its site says it is not affiliated with Nous Research, and the main plan was $49 a month, or $39 a month billed annually, when I checked the official FlyHermes site on 11 October 2026.

Below you'll find what it is, the exact prices, the setup steps, every feature, the limits, a fair comparison with Hermes Cloud, a VPS and your own Mac, and a late section on the models.

I run Hermes every day on my Mac and I've deployed Hermes Cloud, but I haven't run a long hands-on test of a paid FlyHermes workspace.

Everything about FlyHermes on this page comes from its own public pages, and I'll tell you where those pages are vague or disagree.

What FlyHermes is in plain English

Hermes Agent is the free, open-source AI agent from Nous Research.

It keeps memory between sessions, uses real tools, runs jobs on a schedule and talks to you through chat apps.

It needs a machine that stays switched on, a model to think with and a gateway to your chat apps.

FlyHermes is a company that sets all of that up and charges a monthly fee for it.

The site says the plan includes five things.

• The plan includes a hosted server, so the agent doesn't run on your laptop.
• The plan includes model API access for supported providers, so you don't open separate model accounts.
• The plan includes tooling infrastructure for scraping and operator workflows.
• The plan includes core channels, which the home page lists as Telegram, Slack, WhatsApp, Discord and web chat.
• The plan includes a monthly usage allowance, although the home page doesn't state its size.

The headline claim on the site is that web chat is ready in about 60 seconds.

Who makes FlyHermes and is it official?

The site footer says FlyHermes is operated by HF Data Studio LLC, with an address at 447 Broadway in New York.

The refund policy describes HF Data Studio LLC as a Wyoming limited liability company.

The footer says it is an independent service and not affiliated with Nous Research.

Nous Research makes Hermes Agent and runs its own hosted product, which is called Hermes Cloud.

The FlyHermes app page also uses the words "Hermes Cloud" as a heading, so check the web address to see which product you are on.

FlyHermes pricing on 11 October 2026

I read three FlyHermes pages on the same day, and they don't show the same plan list.

The home page shows one plan and labels $49 a launch offer.

The app page at cloud.flyhermes.ai shows three tiers with daily task limits.

The annual prices for Solopreneur and Operator come from the separate Operator page on the main site.

I can't tell you which list you'll see at checkout, so read the plan name and its limits before you pay.

Neither page defines a heavy task or says what happens when the allowance runs out.

How to sign up and deploy FlyHermes

These steps combine the three steps on the home page with the four-step quick start in the FlyHermes docs.

First, you open the pricing section and choose monthly or annual billing.

Second, you click the deploy button and sign in, and the privacy policy says sign-in uses Google OAuth.

Third, you pay at the hosted checkout, which the privacy policy says Stripe processes, and you confirm your workspace.

Fourth, you send a first message in web chat, which the site says is ready first.

Fifth, you connect Telegram, and the docs say you can add your own bot token or invite the managed bot.

Sixth, you connect only the tools you need, and the docs name GitHub, Gmail, Google Workspace, Stripe, Vercel, browser automation and custom APIs.

Seventh, you save stable preferences and rules to memory, which the docs call durable memory.

Eighth, you run one job by hand, check the output and then put it on a schedule.

I'd start on monthly billing, because annual billing charges $468 upfront.

How to use each FlyHermes feature

Channels

The home page lists Telegram, Slack, WhatsApp, Discord and web chat.

The docs call Telegram the fastest first channel and mention email, webhooks and the CLI as later additions.

The Starter tier on the app page lists one connected channel, so the full list depends on your plan.

Memory

The docs describe memory as durable preferences, environment notes and stable project conventions.

They say memory is not for disposable task logs.

Skills

The docs describe skills as reusable playbooks with exact commands, pitfalls and verification steps.

The skills page lists software development, growth marketing, research, productivity, ML operations, and creative and media.

Tools and integrations

The integrations page lists Telegram, Discord, Gmail, GitHub, Google Workspace, browser automation, files and terminal, custom MCP tools and scheduled jobs.

The docs say to store API keys in the workspace environment and never paste tokens into a public thread.

Scheduled jobs

The pages describe recurring digests, monitors, alerts and watchdogs that deliver to a channel.

The home page uses a daily briefing at 7am as its example.

The home plan lists scheduled jobs, and the app page lists them from the Solopreneur tier upwards.

FlyHermes limits and fine print

I read the legal pages, because that's where the real limits are.

• The site doesn't publish server specs, storage sizes or data centre regions.
• The home page doesn't state the size of the monthly usage allowance.
• The terms don't guarantee uninterrupted availability, and the SLA is listed only for team and enterprise plans.
• The terms make you solely responsible for backups.
• The terms say your instance is decommissioned and its data permanently deleted once a cancellation takes effect.
• The terms cap total liability at the amount you paid in the previous three months.

The refund position has two versions.

The refund policy, last updated on 7 June 2026, gives first-time subscribers 14 days to request a refund.

A monthly refund is the first month's payment minus AI credits used, based on FlyHermes usage records.

An annual refund adds the prorated value of the unused months.

The terms on the app domain, last updated on 6 March 2026, still say all payments are final and non-refundable.

The refund policy is the newer page, but I'd get the answer from support in writing before buying a year.

FlyHermes privacy and security notes

The privacy policy says API keys you provide are encrypted at rest with AES-256-GCM, and data in transit uses TLS.

It says card numbers aren't stored and personal data isn't sold to third parties.

It lists Supabase, Stripe, Google and Vercel as third-party services.

It says instance data is permanently deleted within 30 days of cancellation or account deletion.

One line says AI providers are used through your own API keys only, which doesn't match the included model access on the home page.

That policy is dated 6 March 2026, so I read it as an older page that hasn't caught up.

With any hosted agent, your messages and connected accounts pass through someone else's infrastructure, so connect the least you can.

FlyHermes vs Hermes Cloud vs a VPS vs your own Mac

I checked the Hermes Cloud figures on the Nous Portal on 11 October 2026.

Hermes Cloud's Standard size has 2GB of RAM, 2 vCPUs and 10 concurrent sessions.

Thirty days of Standard running non-stop is about $15.30 before model usage, and you need a $2 minimum credit balance or an active Nous subscription to deploy.

FlyHermes costs more per month on paper, and its price includes model access where Hermes Cloud bills inference separately.

A VPS gives you the most control, and my sister post on the best Hermes Agent VPS ranks the hosts and gives the setup commands.

Your own Mac costs nothing extra and is the only route here that can run local models.

My ranking of the best Hermes Agent setup compares the full stacks.

I first covered this service in June in my best Hermes VPS post on the AI Profit Boardroom blog.

Who FlyHermes suits and who should skip it

FlyHermes suits a non-technical owner who wants an agent on Telegram today and prefers one flat bill.

It doesn't suit someone who needs to choose exact models, run local models or know the data region.

It also isn't needed by anyone happy to follow a tutorial, because Hermes itself is free.

My own preference is to own the setup, because you learn how the agent works and nothing disappears when a subscription ends.

Which models sit behind FlyHermes, and how do they score?

This is the part of the decision that GoldieBench can help with.

FlyHermes doesn't ship its own model, and neither does Hermes Agent.

Hermes is the harness, and the brain is whichever model the host connects for you.

The FlyHermes home page names five model families, which are Claude, GPT, Gemini, DeepSeek and Llama.

It doesn't list exact versions, and it says the supported models change over time.

That means I can't tell you which model answers your message on FlyHermes, and I won't attach a score to FlyHermes itself.

What I can show you is how models from those families score on my board.

GoldieBench gives every model the same one-shot prompt to build a game, a page, a simulation or a visual, then scores the result from 0 to 10.

It doesn't measure hosting, uptime or chat quality, so treat these numbers as a signal about the brain and nothing more.

The gap inside one family is the lesson here.

Claude Opus 5 averages 8.27 and Claude Sonnet 5 averages 7.01, so "Claude included" can mean quite different results.

The two DeepSeek V4 models have pages on the board but no average score yet, so I won't give them a number.

There is no Llama model on the board, so I can't score that family at all.

The same family names apply to any host, which is why I'd ask FlyHermes support which exact versions are live before subscribing.

How Hermes itself does with strong models

One entry on the board shows what the Hermes harness can do when the brains behind it are strong.

That run was on my own Hermes install, not on FlyHermes, and I haven't benchmarked any hosted Hermes service.

I don't know whether FlyHermes offers a Mixture of Agents mode, because its pages don't mention one.

What this means when you choose a host

If you pick your own model, on a VPS, on Hermes Cloud or on your Mac, you can choose the exact brain from the board.

If the host picks for you, you get convenience and you give up that choice.

My best LLMs for Hermes Agent post and my ranking of the best Hermes Agent model go deeper on the brains.

If you'd rather run a free brain on your own machine, the local models board shows what runs offline.

Also On Our Network

• Agent Operatorsthe full FlyHermes setup walkthrough for operators
• AI Income Deskhow a business would use a hosted Hermes agent
• AI Tool Verdictevery FlyHermes feature explained and scored
• agentos.guidethe FlyHermes quick-start and how it fits an Agent OS
• agentos.guideMy Hermes Cloud guide on agentos.guide

Start on monthly billing, ask which model versions are live, and judge it on one real job before you commit to a year of FlyHermes.