The Request agent step
This is the handover mechanism inside a workflow. When the script reaches it, the conversation is marked as waiting for a person, an "agent requested" line is added to the thread, and the conversation sits in your inbox unassigned until somebody opens it.
Add it like any other step: click Add a step on the card where the handover should happen and pick Request agent. It has two optional settings:
- Message to the visitor (optional) — what the visitor reads at the moment of handover. It is always sent, unless a colleague is already in the conversation — then nothing is sent. Leave it empty and Yaplet sends its own line, "A team member will join this chat.", in the visitor's language — but only while somebody is available (see below).
- Note for the colleague (optional) — only your team sees it. It is added to the conversation as an internal note just before the handover, so whoever picks the conversation up finds it right above the handover line.
Both can contain values the workflow collected, such as {{ order_id }} or what a Call an API step read. The workflow does not wait for a colleague: it hands over and goes straight on to its next step.
What the visitor sees next
It depends on whether anybody who has access to this brand is available at that moment — switched to Online and not disabled — and Yaplet checks that for real rather than assuming:
- Somebody is available. The visitor gets your message, or "A team member will join this chat." if you left it empty. If nobody has taken the conversation or replied 5 minutes later, the website chat shows a "Leave a message" card titled "The team hasn't been able to reply yet" — unless the brand has its own switched-on Nobody picked up workflow (see below).
- Nobody is available. In the website chat, the "Leave a message" card appears right away, titled "The team is away right now". The visitor writes what they need right in the chat and can leave an email address, so your team can also answer by email; if Yaplet already has their address, it is filled in. Where the AI can still answer, the card also offers to keep chatting with the AI. On Facebook and Instagram, which cannot show cards, the visitor gets a short built-in "we are not available right now" notice instead.
Everything the visitor is shown is stored in the conversation, so a colleague opening it afterwards sees exactly what the visitor was told. None of the built-in texts can be edited anywhere in Yaplet — each visitor reads them in their own language.
What your colleague sees
- The whole scripted conversation, in order — every message the bot sent, every button the visitor clicked, and every answer they typed.
- What you wrote under Note for the colleague, if anything, as an internal note the visitor never sees.
- The visitor's email address on their profile, if the script collected one.
- The same thread, continued. The visitor never has to repeat themselves.
Who is allowed to pick it up
Access is decided per brand. A teammate can only see a conversation if they have been granted the brand that the chat widget belongs to. Owners and administrators always have access to every brand; everyone else is granted brands explicitly.
- Open your account menu — your avatar in the top right — and choose Manage team.
- Open the teammate.
- Under Brand access, switch on the brands they should be able to work in, and save.
This also decides the "is anybody available?" check above: only people who can actually open the conversation are counted as available.
Setting expectations before the handover
Use Message to the visitor to say what happens next in your own words — it is shown whether or not anybody is available, so write it to fit both cases. Yaplet never promises a reply time. An older workflow may still have a Reply time (retired) step before Request agent: it does nothing and you can delete it.
When nobody picks up
You can also plan for the moment nobody takes the conversation. Build a separate workflow and choose Nobody picked up under Starts when. It starts the set number of Minutes (5 by default, 45 at most) after the visitor asked for a colleague — but only if nobody has taken the conversation and nobody on your team has written to the visitor since. It runs once per request; if the visitor asks again, the clock starts over.
Use it to apologize for the wait, ask for an email address so your team can answer later, or offer a callback. While such a workflow is switched on, it takes the place of the built-in card the website chat shows after 5 minutes. It works in website chat, Facebook, Instagram and email — not on phone calls, where nobody is ever handed over.
Handing over to the AI instead
If what you want is an AI answer rather than a person, use Send to the AI agent in place of Request agent. You write the question that gets passed over, and the brand's Yaplet agent answers it from everything the brand knows. If the brand has no Yaplet, or Yaplet does not answer on that conversation's channel (its Answers on switches, on Brand → AI agent → General), the step hands the conversation to your team instead.
There is a related setting worth knowing about on the AI side. On Brand → AI agent → Human hand-off, in the Ways to ask for a person group, there is a switch called Visitor asks Yaplet for a person. With it off, Yaplet never connects anyone itself: when a visitor asks for a person, it points them to Request agent in the ⋮ menu at the top of the chat window, or on Facebook, Instagram and email to the contact details it finds in the brand's knowledge. If you would rather handle those requests your own way, the screen points you to the recipe — build a workflow for this brand with the trigger User says and the intent "The user wants to talk to the support team", and let that workflow do the handing over.
One channel note
A Request agent step rules the whole workflow out of phone calls. On a call there is nobody to hand the caller to, so a script containing this step is simply never offered on the phone. It works normally in chat, in Facebook and Instagram messages, and in email conversations.
Next step
Learn how several workflows behave when they could all fire at once: Run multiple workflows at the same time. Not sure whether to script this at all? Which chatbot should I use — Yaplet or workflows?