Ask us anything.
Here are the answers.

No. If you can use email, you can use this. You press one button, we build the server and set the worker up for you, and the next thing you see is a chat window. Nothing to install, nothing to configure, nothing to keep updated.

About three minutes from signing up to sending your first message. Most of that is the server being built, which happens while you watch a progress bar.

Say hello to it. Genuinely — ask it what it can do, or give it something small and annoying like tidying your inbox. You do not have to plan anything or write instructions. It is a conversation.

No. You can use it in your browser, or connect it to something you already use — Telegram, a plain text message, Slack, Discord — and talk to it there. No new icon on your home screen.

There is no free trial, but there is a 30-day money-back guarantee, which amounts to the same thing with less faff. Start on Basic at $19.99/mo, and if it is not for you, say so and we refund the plan.

See what each plan costs

You give it one when you create it. It answers to whatever you pick.

That is completely reasonable and it is why we wrote a whole page in plain English — what AI actually is, what makes this a worker rather than a chat, what a private server is, and how the money works. No jargon, with pictures.

Read the plain-English explainer

Read and answer email, manage your calendar, set reminders, keep lists, research things and come back with the short version, read and file documents, and act inside any app you connect it to. The everyday version is: it does the admin half of your day so you can do the other half.

A chat can only produce words — you ask, it answers, and then you go and do the work. A worker has hands. It opens your inbox and replies, puts the thing in your calendar, adds the row to the spreadsheet. You are not getting advice about your inbox, you are getting your inbox handled.

That depends on which worker you choose and what you allow. Hermes checks with you before the things worth checking — sending on your behalf, spending, deleting — and does the harmless things quietly. You can tighten or loosen that. If the idea of it acting alone makes you uneasy, keep the approvals on.

The difference between the two workers

Yes. It lives on a server that never switches off, not on your laptop. Set something going at midnight and it is done by morning. It also does not take lunch breaks, weekends, or offence.

Yes, and that is most of the point. It keeps a short note about you and how you like things done, and a library of jobs it has already worked out, so the second time you ask for something you barely have to ask.

How the memory works

Yes, and it will sound just as confident when it does. That is the one thing worth remembering about all AI. Treat it like a fast, very well-read colleague who occasionally misremembers — useful constantly, worth checking on anything that really matters.

A worker belongs to one seat. If you want one for a colleague, add another seat — each gets its own server, its own connected accounts and its own memory, which is usually what you want anyway. If you are looking at a lot of seats, talk to us rather than adding them one at a time.

Many seats and enterprise

How you like to be written to, who the important people are, the rules you keep repeating, and the way your setup works. Small, useful things — not a transcript of your life. It writes them down as it goes, without you having to remember to tell it.

Yes, and you can change it or delete it. It is two short, plain text files sitting on your own server — one about you, one about your work — not a hidden profile somewhere. If it has written something down that is wrong, cross it out.

Exactly how the memory works

A written-down method for doing something. When your worker figures out a fiddly job — chasing unpaid invoices, writing the Monday report — it saves how it did it. Next time it just does it, the same way, without being talked through it again.

Three places. It writes its own as it works. You can install ready-made ones. And you can point it at some documentation and say learn this, and it will write the skill itself. They live on your server and they are yours.

Installing and writing skills

That is the idea, and it is why this is a subscription rather than a thing you buy once. Month one you are explaining yourself. Month three it already knows. The moment you notice is usually the day you catch yourself not repeating an instruction.

Yes. It lives on a server that never sleeps, so it is never away and never offline. Send it something between jobs, at the end of a long day, or at eleven at night, and it starts working — you do not have to be at a desk, or even awake, for it to be getting on with things.

No. The notes and the skills stay on your server. Even cancelling only parks your worker rather than deleting it, so if you come back everything it learned is still there.

Telegram, plain text messages, Slack, Discord and WhatsApp, plus the chat window in your own dashboard — and more get added. Connect a channel once and your worker turns up there as an ordinary conversation.

Yes. If your hands are full, hold the button down and talk. It listens and gets on with it, exactly as if you had typed. Useful when typing out the whole thing would take longer than saying it.

No. It is one worker with one memory. Ask it something from your phone while you are out, open the dashboard when you are back at your desk, and it knows precisely what you were talking about.

No. Lionox is an independent service and is not affiliated with WhatsApp, Meta, Google, or any messaging platform. You connect your own accounts, and you can disconnect any of them in one click.

If you are technical, yes. Bring your own WhatsApp Business number and credentials, register your own Telegram, Slack or Discord app, and point the webhooks wherever you need them. The whole conversation path can sit inside accounts that belong to you.

How to run your own connectors

Over 1,167 of them — Gmail, Outlook, Google Calendar, Drive, Sheets, Notion, Slack, HubSpot, Salesforce, Stripe, Shopify, Xero, QuickBooks and a very long tail after that. You can search the whole list and see exactly what your worker could do inside each one.

Search your tools

No, and we do not want them. Connecting an app opens that app's own sign-in page, you approve it there the way you would with anything else, and it hands back permission rather than a password. We never see your password.

One click and a sign-in. There are no API keys to copy, no settings to fill in and no developer account to create. Every app connects the same way, so once you have done one you have done all of them.

Yes, in one click, and the ability goes away immediately. Nothing about connecting an app is permanent.

Tell us and we will look at it — the catalogue grows every week. In the meantime a worker can also read a plain web address, or use an API you point it at, so quite often it can get there anyway.

No. It is useful the moment it exists — ask it questions, get it to research, have it draft things. Connecting apps is what turns it from clever into genuinely useful, but there is no rush.

They are the same idea — a helper that lives on a server and goes and does things — but built with different attitudes to risk. OpenClaw came first, is open source, and is built to go off and act on its own. Hermes was designed from the start for a worker acting on someone else's behalf, so the permission checks and approvals are part of its structure rather than bolted on afterwards.

Hermes. It is the default and the recommendation, and it comes from Nous Research. You lose very little and you gain a worker that asks before doing something you would rather it asked about.

If you want maximum independence and you are happy supervising it. It is genuinely good at long, messy jobs with lots of browsing and files, and it has an enormous community behind it. Give it only the connections it needs, start it on something low-stakes, and check its work on anything that matters. If that sounds like effort, use Hermes — nobody will judge you.

Yes. The worker type is a setting, not a life choice. If one is not working out, switch.

The explainer page covers both in plain English, alongside what AI is, what a private server is, and how tokens and credits work — with diagrams rather than paragraphs.

Hermes and OpenClaw, explained

A model is the AI brain your worker thinks with. You do not have to care — it comes set up sensibly and you never have to touch it. It matters if you want to spend less, or if you have a job one particular brain is better at.

ChatGPT, tuned for Hermes. It is a good all-rounder and there is no reason to move it unless you want to.

A couple of dozen well-known ones, picked per conversation. Broadly: Claude is the strongest reasoner, Gemini is the best with pictures, and DeepSeek is extremely cheap and fast. The price of each is shown right beside it when you choose, so you always know what you are about to spend.

See what each model is good at

The same idea in miniature — how hard it thinks about a given job. Light for sorting and deciding, Balanced for writing, Strong for the long careful jobs. The difference is not small: Strong costs roughly Dearest what Light does, so leaving everything on the heaviest setting is the fastest way to burn through credits for no benefit.

Yes. Point your worker at your own provider account and you pay that provider directly at your own rates, on your own bill and your own rate limits. You can mix it — your key for the heavy models, our credits for everything else.

Bring your own key

Yes. Install Ollama on a machine you own, pull an open-weight model onto it, point your worker at the endpoint, and the thinking happens on your hardware. No credits are spent on those calls and no prompt reaches a third-party model provider. Worth being precise: your worker still builds the prompt on its own instance first, so this gives you model independence rather than full data residency.

Running your own model

Two parts. A monthly price for the plan — your own private server and the worker running on it — starting at $19.99/mo. Then AI credits, which are spent when your worker thinks, searches or acts in an app. Every plan includes free credits each month, from $2 on Basic up to $30 on Max.

All four plans side by side

AI reads and writes in tokens, which are word-sized chunks — roughly a hundred words to a hundred and thirty tokens. Tokens cost money, and credits are simply your balance for them. For everyday use this rarely comes up: you leave the defaults alone and you rarely think about it again.

Tokens and credits, explained simply

You get a warning at 80% used, both on the billing page and next to your worker. It carries on answering until the balance reaches zero, then waits until you top it up. It does not silently start charging you.

The free monthly credits reset at the start of each billing period and do not roll over. Credits you buy on top are different — those never expire, and they are only spent once the month's free ones are gone.

It is on by default for a new worker so it never goes quiet mid-job: your card is charged $10 whenever that worker's balance drops below $1, capped at $50 a month. The setting and the cap are per worker rather than per account, so two workers left on the defaults can reach $100 between them. You can change the amount, change the cap, or switch it off entirely in Billing.

Yes — 5%, whether you top up yourself or leave auto top-up on. Pay $10 and $9.52 of credit lands on your worker. That is the card fee and the handling, said out loud rather than buried in a footer.

Because you choose the model and the thinking depth, and those differ enormously in price. There is also a rule of thumb worth knowing: reading is cheap and writing is expensive. Skimming a hundred emails costs very little; producing a long polished document costs a great deal more.

Start on Basic. The bigger plans buy you a faster machine and more included credits, not extra abilities — every plan can do everything. Move up when you notice you are topping up before the month is out.

The same machine — Max is not faster. What it adds is twice the storage, twice the credits, and we set it up with you one to one. Worth it once you are topping up Pro every month.

Yes, at any time, from Manage billing in your dashboard. Change tier, update your card, or cancel — two clicks, no phone call, no retention conversation.

No setup fee, no contract, no minimum term. It is a monthly subscription you can stop whenever you like.

No. Your server is yours alone — your worker, your files and your conversations live in their own locked space, and no other customer shares it. That is the point of giving everybody their own instead of a seat in a shared pool.

It is a computer in a data centre that never gets switched off, running only your worker. The jargon for it is a VPS; the important word is private. It is also why your worker is always on — it is not running on your laptop, so it does not stop when your laptop does.

Private servers, explained simply

No. You never see it, never log into it, never update it. We build it, run it and keep it patched. You only ever see the chat.

Yes. Ask us for a full deletion and we remove the lot — server, files, conversations. That is a deliberate, separate request rather than something that happens by accident when you cancel.

Three things: the approval checks, which pause it before the actions worth pausing on; the connections, since it can only touch apps you have linked and you can unlink any of them instantly; and the isolation, since it only has your data to work with in the first place.

Then you get the plan fee back. Tell us within 30 days of your first payment and we refund it in full — no forms, no phone call, no reason needed. Credits your worker already spent are not refundable, because that work has been done.

The refund covers your plan — the server and the subscription. AI credits your worker has already spent are not refundable, because that work has already been done. Unspent credits you have bought are not refunded either, but they never expire, so they are still there if you come back.

It goes to sleep rather than disappearing. Cancelling parks the instance and keeps your files and conversations, so if you come back later everything is where you left it. If you would rather it was all gone for good, ask for a full deletion and we will remove the lot.

Yes. Cancel a seat from Manage billing in your dashboard whenever you like. You will not be charged again, and there is nothing to send back.

Then ask us — a real person answers, and questions people actually ask are how this page gets better.