Lists your twin can look up

Most of what your twin knows is prose: your writing, your bio, how you explain things. But some things you keep as a list — what you offer, roles you're hiring for, products you sell, questions you answer over and over.

Your twin treats those differently: it looks them up and quotes them, rather than describing them from memory. Ask "what do you charge for a pricing review?" and it reads the row, not a paragraph that happens to mention pricing.

You'll find this in the Lists tab.

Starting a list#

Pick a starting point — Open positions, Services, Products, Workshops, Resources, Common questions, Known issues — or start from nothing and name your own fields. Each starting point is a real list you then edit, not a blank form.

Every list has two names, because your twin uses both: what one of them is called ("Open position") and what several are called ("Open positions").

The sentence your twin reads#

Each list has a one-line description, and this one isn't a label for you — it goes to your twin before every answer, and it's how your twin knows what the list is for.

Roles currently open, with where they are and how to apply.

That's useful. "Positions" isn't. A vague description gets you a vaguer twin.

What each entry holds#

Three things are built in, on every entry of every list:

Titlerequiredshort text, up to 120 characters#

What the entry is called. It counts three times as heavily as anything else when your twin works out which row answers a question, so name the thing rather than describing it.

Descriptionoptionallong text, up to 600 characters#

The sentence or two your twin quotes back. An entry without one still works — it just has less to say.

Linkoptionallink, starting http:// or https://#

Where the entry points. Mark it as a booking link and the entry gets a Book a time button on your own twin — never on a Support Agent or a Team Agent. See Booking and applying.

Add your own fields for anything else you keep: a price, a location, a closing date, whether something is in stock.

Each field holds one kind of thing — short text, long text, a number, yes/no, a date, a link, or a choice from a list you set.

"Your twin can narrow by this"#

Tick this on a field and your twin can filter by it. Tick it on Location and someone can ask "any roles in Berlin?" and get exactly those.

You'll notice the tick isn't offered on free-text fields. That's deliberate: filtering needs exact values, and a promise your twin can't keep precisely is worse than no promise. It can still read and quote a text field — it just won't claim to have sorted by it.

Filters match exact values today. "Roles under £60k" isn't something your twin can narrow by — but the salary is on each row, so it can show you the Berlin roles and let you read the numbers.

Prices stay exactly as you type them#

"$500", "from £2,000", "DOE", "45 min" — all fine, and all shown exactly as written. We never compute, convert, or estimate a price on your behalf. Same rule that governs everything else your twin says.

Preparing something before it's live#

Each entry has "your twin can use this". Untick it and the entry is invisible to your twin — so you can write a role up before it's announced without a candidate hearing about it early.

Booking and applying#

Mark a link field as a booking link and the entry gets a Book a time button — on your own twin. A Support Agent or a Team Agent never shows it; the end of this section says what they show instead.

• Cal.com and Calendly links open inside the conversation, in a sheet, so the visitor never leaves your twin. The sheet has a ✕ in its corner and also closes on a tap on the dimmed area above or below it; what a visitor sees says exactly when Escape works and when it doesn't, because the booking form itself is the scheduling service's own page.
• Any other link opens in a new tab, because we can't guarantee another site allows being embedded.
• Opening the booking sheet counts as a booking click in your Insights, the same as following a link out. Reopening the same sheet in one conversation isn't counted twice.

You don't need to connect a calendar or authorise anything: paste the scheduling link you already use.

A Support Agent or a Team Agent never shows the button. Those two kinds of agent can't offer what you sell — that's the promise on the Support Agent page — so a booking link is withheld from them: no button, and the address itself is kept out of the answer too, rather than left where the agent could paste it into a sentence. The rest of the entry is unaffected: the title, the description and every other field still answer questions, prices included. Your own twin shows the button as normal.

When your twin brings a list up#

Only when it genuinely answers what was asked. Ask about the weather and it won't offer you a call.

Your twin quotes only what a lookup actually returned. If nothing in a list matches, it says so — it won't describe what it expects you'd have, and it won't invent a price, a location or an availability you never entered. If a list is long, it says these are some of them rather than implying it read every one.

Sharing a list between twins#

A list belongs to your account, and each twin carries the lists you give it. If you later have more than one twin, they can share a list — fix a price once and every twin quoting it is correct.

Detach from this agent doesn't delete anything. The entries stay, so you can put it back later exactly as it was — the same button reads Add to this agent once it is off, and it is the undo, one tap away in the same row. That is why it asks for no confirmation: nothing is lost. Useful if you're hiring some months and not others.

On your own twin the same two controls read Detach from this twin and Add to this twin, and a list you haven't given it is marked Not on this twin. The page calls the thing a list belongs to by the name the rest of your console uses for it — agent on a Support Agent or a Team Agent, twin on your own.

There is no button here that deletes a list, and that is deliberate rather than missing: detaching is the reversible act, and a list your account keeps costs nothing while it sits unattached. If you want one gone for good, ask us.

Keeping a list current automatically#

Some lists you write once and edit occasionally — services, open roles. Others change faster than anyone will keep up with by hand: stock, prices, a catalogue of hundreds.

For those, point a list at a URL that returns your rows — JSON or CSV, which is what most shop exports, ATS exports and published spreadsheets already give you. Say which incoming column becomes which of your fields, pick hourly or daily, and the list refills itself.

Two things worth knowing:

• Anything you typed by hand stays. A sync only ever touches the rows it created, so your own entries in the same list survive a feed that doesn't mention them.
• Columns you didn't map are ignored, never guessed at. If your feed has an internal margin or a warehouse code, your twin never sees it, because guessing is how a warehouse code ends up quoted to a customer.

A feed that returns more than 500 rows is read down to the first 500, and the sync tells you how many it left behind.

If a row doesn't fit your fields — a category that isn't one of your options, a word where you asked for a number — it's skipped and reported, rather than stored as something your twin would then state as fact.

Limits#

Those aren't arbitrary: your twin is told about every list and every filter before it answers, so an unlimited setup would make every single answer more expensive.

A note on tone#

The point of a twin that admits what it doesn't know is that people trust what it does say. A twin that constantly steers toward a sale spends that trust. The defaults here are deliberately conservative for that reason — and a list your twin shouldn't mention is one you simply don't give it.