Getting started
Your first 15 minutes with Unyo: create an account with 2,500 free credits, meet your agents in a 7-screen onboarding, fill your Neural Core, connect one integration, and run a real task.
Questa pagina non è ancora disponibile nella tua lingua - stiamo mostrando la versione in inglese.
Fifteen minutes from now you'll have an account, a team of agents that knows your company, one connected tool, and a finished piece of real work. Here's the whole sequence, in order.
You don't need a credit card, an API key, or a connected account to begin. You need your company name and about a quarter of an hour.
1. Create your account (2 minutes)
Go to the sign-in screen and pick one of three routes:
- Continue with Google
- Continue with Apple
- Email and password - you'll enter a first name, last name, email and a password of at least 6 characters.
That's the whole form. We don't ask for your company at signup - Ema asks you on the very first onboarding screen instead, because it's a better first conversation than a form field.
2,500 free credits, no card
Every new account starts with 2,500 credits and no payment method on file. Credits are a real unit. And because 73 of our 105 priced tools cost zero credits, your free balance goes much further than it sounds. Every draft and every plan is free, and so is almost every read and search - the exception is your calendar, where each lookup costs 10 credits. You're only charged when something is actually produced.
If your browser is set to French, Spanish, German, Turkish or Arabic, the sign-in screen appears in that language automatically. The app itself runs in English, French, Spanish, German, Italian or Turkish - you'll pick yours in step 5.
2. Meet your team - the 7-screen onboarding (4 minutes)
The first time you sign in, you land in onboarding rather than the dashboard. It is seven screens, and each agent asks you for exactly one thing. No form, no wizard with 12 fields. One question per screen.
| Screen | Who asks | What they want |
|---|---|---|
| 1 | Ema | Your company name |
| 2 | - | A short welcome - what the team is |
| 3 | Maya | A description of what your business does |
| 4 | Riley | Your website URL |
| 5 | Tyron | Your logo |
| 6 | Ashley | Which tools you want to connect |
| 7 | - | You're ready |
This is not decoration. Everything you type here lands in your Neural Core - the shared memory that all 10 agents read on every single run. The company name Ema takes on screen 1 is the same company name Lucy uses in a press release on day 30.
Answer screen 3 properly
The description you give Maya is the single highest-leverage thing you type all week. "We sell things online" produces generic output from every agent, forever. "We're a two-person Paris pâtisserie doing wedding cakes and corporate orders, mostly Instagram-driven, average order €180" produces work you can actually use. Spend 60 seconds on it.
You can skip a screen and come back - everything on it is editable later in your Neural Core and in Settings.
3. Fill your Neural Core and your Brand (4 minutes)
Open the Neural Core from the sidebar. This is your business memory, and it's the reason Unyo gets better instead of staying flat.
You'll see a three-card hero - Products, Contacts, Files - plus Rules, Knowledge and Tasks.
Do these three things now:
- Drop in one real file. A price list, a product catalogue, a brand PDF, a CSV of customers. Up to 50 MB per file (20 MB for video). Images, video, PDF, CSV, XLSX, XLS, ODS, TSV, TXT, MD, DOCX, PPTX, JSON and SVG are all accepted. Images and videos get described automatically; a messy CSV dropped onto Contacts or Products is parsed into real entries - the separator is auto-detected and the headers are matched even if they're in French, German, Arabic or Japanese.
- Write one rule. Go to Rules and add something like "Never use emojis in emails" or "Always sign off as the team, never as an individual." Every agent follows it from that moment on - you don't repeat it to each one.
- Set your Brand. Open Settings → Brand and enter your identity, slogan, sector, contact details, socials, logos (light and dark), colours, font and tone. All 10 agents read the same brand. You set it once.
Your Brand is read-only to the agents
Agents don't manage your Brand - you do. The save path, the update and delete tools, and the auto-learner all refuse the brand category outright, in code rather than by asking the model nicely. The same applies to products synced from Shopify: they're your source of truth, so every write tool rejects them. If you want your brand changed, you change it.
The Neural Core is free
Every Neural Core tool costs 0 credits - reading it, searching it, writing to it, exporting it. Memory is not a metered upsell. You're only ever billed for the run itself, at real cost.
You never have to fill this in by hand again. From here on, agents learn as you talk: tell Maya "our new price is €49" in the middle of an email conversation and it's saved - no command, no /remember prefix, in any language. And anything you delete stays deleted: deletion writes a tombstone that's fed back into the learner as "do not re-extract this".
4. Connect your first integration (2 minutes)
Open Integrations from the sidebar (it's a top-level item - not inside Settings).
Sixteen integrations are live: Gmail, Outlook, Google Calendar, Google Drive, Instagram, Facebook, LinkedIn, LinkedIn Page, X, Slack, Trello, Notion, Shopify, WooCommerce, PrestaShop and Stripe.
Connect one - the one that matches the work you actually do:
- Email is your day → Gmail or Outlook (see Gmail / Outlook)
- You post → Instagram, Facebook, LinkedIn or X
- You sell → Shopify, WooCommerce or PrestaShop
- You want numbers → Stripe
Only Maya has your inbox
When you connect Gmail or Outlook, exactly one agent gets access: Maya. Ashley cannot read your email. Nor can Blake, Lucy, Steve or anyone else. This isn't a policy in a prompt - the email tools are denied to the other nine agents at the runtime layer, so the request cannot be made in the first place. Every agent boundary in Unyo works this way.
Note also: one Gmail connection per account. You can connect Gmail and Outlook, but not two Gmail accounts.
You don't have to connect anything to start. If an agent needs a tool you haven't linked, it doesn't fail and it doesn't guess - it shows you a Connect card, pauses, and resumes the moment the connection lands. You won't retype your request.
5. Set your languages (30 seconds)
Two settings, deliberately separate, both in the Settings modal (open it from the sidebar on desktop, or the tab bar on mobile):
- Interface language - English, French, Spanish, German, Italian or Turkish. Changes the app.
- Agent language - 11 options. Changes what your agents say, and nothing else.
So you can run the interface in English while Maya writes your customers in Portuguese. Voice supports 11 languages.
6. Run your first real task (3 minutes)
Go to Home, use the input bar, or open a conversation with an agent directly. Pick the one that matches your step 4 connection.
If you connected an inbox - ask Maya:
Then, on anything worth answering:
You'll get an editable draft card in the chat. You read it, change it, and press send. She was never going to send it without you.
If you connected a social account - ask Ashley:
Write me an Instagram post announcing our new opening hours. Give me the caption, the hashtags, and generate an image for it.
You get a draft card with a caption, hashtags and a generated image, in a side panel you can resize. Edit it, then publish it yourself.
If you connected nothing - ask Steve to build you a site:
Build me a one-page site for my business: a hero, three services, and a contact section. Use my brand colours.
You'll watch the code get typed into a live file tree as it's written, then switch to a Preview tab and click any element to tell him what to change.
What no auto-send actually means
Unyo drafts. You send. This holds everywhere - email, social posts, publishing a website - and it holds for scheduled work too: an automation prepares a draft and stops. There is no auto-send toggle to find, because unattended sending is refused in two separate places in the code. It's the honest description of the product and, we think, the better promise.
What you'll notice while it works
- The agent shows its work. It declares its plan, then each step flips from pending to in-progress to done or failed, live. A failure stays visible with an amber ✗ - it isn't quietly swallowed.
- You can expand any step to read what the agent actually found, in your interface language.
- You can close the tab. The run keeps going on the server. Come back and the finished work is in the conversation.
- Refresh mid-run and nothing breaks. What you see streaming and what you see after a reload are produced by the same code path, so they don't drift.
Your next 15 minutes
Once the first task lands, these are the highest-value next moves:
- Ask another agent about the first one's work. The last 18 hours of activity across every conversation and every agent is shared, so it will already know.
- Ask an agent for something outside its job. It hands the work to the right coworker automatically - the coworker opens its own conversation, does the job, and notifies you at hand-off. (Steve is the exception: to build a website, open a Steve conversation yourself.)
- Ask for something recurring. "Every morning at 7, prepare a summary of what needs me." The agent creates a real automation that runs on schedule and leaves you a draft.
- Read Credits and billing so you know what's free (most of it) and what isn't.
Automations are created by asking, not by clicking
A new account starts with zero automations. There is no template library to switch on. You get one by telling an agent what you want, in words. Once it exists you can pause, resume or delete it from the Automations hub - but creating or editing one means talking to an agent.
FAQ
Do I need a credit card to start?
No. You get 2,500 credits at signup with no payment method on file, and nothing asks for one until you choose to add credits. Drafts, plans and almost every read and search are free, so most of what you'll do in your first session costs nothing at all.
Do I have to connect anything to try it?
No. You can chat with all 10 agents, fill your Neural Core, write documents, build spreadsheets, and have Steve build and publish a website without connecting a single account. Integrations unlock the work that needs your data - your inbox, your calendar, your channels, your orders.
Can an agent send an email or publish a post without me?
No. Agents draft; you send. There's no auto-send setting, and unattended sending is blocked in code rather than discouraged in a prompt. This applies to scheduled and background work too - an automation prepares the draft and stops there.
Do I have to repeat myself to each agent?
No, and this is the part most people don't expect. All 10 agents read the same Neural Core on every run - the agent id isn't even a factor in what gets loaded. Tell Maya "no emojis" and Ashley knows it on her next run. Tell Ema your company name in onboarding and Lucy uses it in a press release a month later.
What if I skip the onboarding questions?
Nothing breaks - you land in the app and the agents work. They just work generically. Everything the 7 screens asked for is editable afterwards in your Neural Core and in Settings → Brand, so the fastest fix is to go fill those in. The description and the brand are what move the output most.
Where do I change my settings?
Settings is a modal, not a page - open it from the sidebar on desktop or the tab bar on mobile. It has five sections: Account, Preferences, Automations, Subscription and Credits. (Your Brand lives under Settings too.) Integrations, however, is its own top-level item in the sidebar.
Troubleshooting
The onboarding didn't appear. It only runs once, on the first sign-in of a new account. If you signed in and went straight to Home, everything it would have asked for is available in your Neural Core and in Settings → Brand.
An OAuth connection failed. Allow pop-ups for the sign-in window, then try again from Integrations. If it still fails, disconnect and reconnect.
An agent stopped and showed me a Connect card. That's intended - it needed a tool you haven't linked. Connect it and the run resumes on its own from exactly where it stopped. If you connect it much later (more than 24 hours), the run won't auto-resume - just ask again.
An agent seems to be ignoring my brand. Check Settings → Brand is actually filled in, and check the entry isn't marked Private in your Neural Core. A private entry is deliberately excluded from anything an agent generates - emails, posts, documents.
Something failed and I was charged. You weren't. Every paid tool is refundable on error, and an hourly job refunds anything stranded. If your balance looks wrong, check Settings → Credits.
My file wouldn't upload to the Neural Core. The caps are 50 MB per file, and 20 MB for video. Chat attachments are capped at 20 MB per file.