Tone is calm and warm. No exclamation marks.
From Homepage and About pageBrain
Explain your brandfor the last time.
The Brain learns your brand, your taste and how you differ from your competitors, and gets sharper every time your team approves or rejects a result. Every agent, teammate and assistant already knows.
Start free100 free credits a month. No credit card.
Three ways in. No forms.
Point it at your site, drop in a document, or paste what your assistant already knows.
Reading amberlane.co
- Palette: terracotta, linen, inkfrom amberlane.co
- Type: Fraunces and Geistfrom amberlane.co
- Voice: calm, warm, never shoutyfrom amberlane.co
- Hero product: Terracotta Candlefrom amberlane.co
It gets smarter.
Most tools forget you the moment you close the tab. The Brain keeps learning, so the hundredth brief needs less explaining than the first.
Point it at your site. It reads your palette, fonts, voice, products and what you claim, and keeps your real product photos.
- LookInk, sand and one caramel accent
- VoiceShort sentences. Never says “luxury”
- ProductsFive fragrances, all vegan
Example memories for a demo fragrance brand.
Nothing goes live unread.
Anything the Brain worked out on its own waits for a person. Try it: one of these facts is wrong.
Needs review · 5Very sure Very sure Unsure
Never show discounts in red.
From Brand book, page 12Brand color: bright red #E3171F
From Sale banner on /shopIn the Brain · 0
- Confirmed facts land here, each with its source and who signed off.
Your brand, in five tabs.
Everything the Brain knows sits in one profile: who you are, how you look, how you sound, the rules and the market. Names you mark Keep are never translated.
- Name
- A demo fragrance brand
- What it sells
- Five vegan fragrances in refillable glass
- Keep
- Salt Hour, Dune Linen (product names, never translated)
Example profile for a demo fragrance brand.
Rules your agents actually follow.
Mark a word Never, Preferred or Careful. A model follows a short rule far more reliably than a paragraph of guidelines.
An illustration of how rules steer an agent’s copy. In the app, rules apply to every agent and assistant that reads your Brain.
Taught once, there everywhere.
The same Brain sits behind every way you make things in Orisu, and behind your own assistant.
The copilot
Starts every chat knowing your identity and word rules, so launch copy comes back in your voice.
The canvas
The Brain node hands your rules and references to any step of a workflow.
The Feed
Every row in a bulk run carries the same brand, product after product.
Claude and ChatGPT
Your own assistant can read and search the Brain. What it proposes lands in your review queue.
The specifics.
How the Brain keeps itself honest and current, so you can trust what it tells your agents.
- Every fact cites its source
Facts from a document point back to the page they came from. Facts from your site point to the page they were read on.
- Who taught it, who signed off
Each memory shows the person or agent that added it and the person who confirmed it.
- Memories expire on purpose
Channels and feedback last about 90 days, campaigns about six months, products and audiences about a year. Identity and standing rules never expire.
- It asks, briefly
When it spots a gap it asks in the copilot, never more than two questions in one conversation.
- The Brain chip on Create
Turn it on and your prompt is rewritten in your brand before it runs, for a few credits.
- Map, Memories, Media
See the brand as a map, as a list of memories, or as the product images it keeps.
- Off for one chat
The copilot uses the Brain by default. A switch turns it off when you want a blank slate.
- One brand per workspace
Facts never leak between brands. Agencies keep each client in its own workspace.
Before you ask.
The rest is in the Brain guide.
Where does the Brain get its facts?
From three places: your website, documents you upload, and memory pasted from an assistant like ChatGPT or Claude. Every fact keeps its source.
Will it use something I haven't approved?
Anything Orisu inferred on its own arrives unconfirmed and waits in Needs review until a person confirms, edits or dismisses it.
Does it learn from what we approve?
Yes. When your team approves, rejects or edits results in a review step, the Brain looks for patterns. A repeated one becomes a preference your agents follow next time; a one-off click is ignored.
Can I correct a fact?
Yes. Edit it in place. A correction counts as a confirmation, because a person wrote the words.
How many brands can I keep?
One brand per workspace, on purpose, so facts never leak between brands. Pro and Enterprise plans include more workspaces.
How long does a memory last?
It depends on the kind. Channel and feedback memories last about 90 days, campaigns about six months, products and audiences about a year. Identity and standing rules never expire. An expired memory needs a Still true from a person to count again.
Will it keep asking me questions?
Only about gaps, and never more than two per conversation. Answers are saved, so it doesn't ask the same thing twice.
Can I turn it off for one chat?
Yes. The copilot uses the Brain by default, and a switch turns it off for that conversation.
Can my assistant in Claude or ChatGPT add facts?
It can read and search the Brain, and anything it proposes lands in your review queue first.
Is my brand data used to train models?
No. Your Brain is private to your workspace and never used to train models.
Teach it your brand today.
Paste your URL, review what it learned, and run your first on-brand agent. Free, no credit card.
Start free