Support chatbot for online stores

Your support knowledge, answering customers at 3am.

Buddy reads your help centre and your order data, answers in your customer's language, and hands over to a human the moment it should.

Two lines on your site

<script src="https://widget.support.bdy.app/widget.js"></script>
<script>SupportBuddy.Init('my-store')</script>

Thirty seconds

See Buddy on a real storefront

One resolution, start to finish

What a good answer looks like

Buddy checks the real order before it promises anything, and steps aside the moment the conversation needs a person.

  1. I ordered the wrong size

  2. I can see order #40219 — a 38 in the navy. Shall I start an exchange for a 40?

  3. yes please, and it arrived damaged

  4. Mette from support joined
  5. I'm taking this one over — we'll replace it and refund the postage.

Now go and break it. Our own bot is in the corner of this page, running on our own help centre. Ask it something we haven't scripted.

What it actually does

Three things, done properly

It answers from your help centre

Your articles are chunked, embedded and linked into a knowledge graph, so every answer traces back to the policy it came from instead of to the open internet.

Retrieved for this question

  • Returns and exchanges
  • Damaged on arrival
  • Delivery times by country

And from your live orders

The WooCommerce connector looks up the real order before Buddy promises anything about it — the status, the items, the address it is actually going to.

Order #40219

  • Navy jacket, size 38
  • Shipped 11 Sep, out for delivery
  • Copenhagen, DK

And it knows when to stop

Escalation follows the rules you write, and only fires when a human is actually online to take it. When nobody is, Buddy says so rather than inventing an answer.

Handing over

  • Rule matched: damaged item
  • Support online: yes
  • Mette from support joined

Pricing

Priced per conversation, not per resolution

Nobody has to argue about whether something counted as resolved, and the handoff to a human never costs extra.

Starter

€49 per month

One storefront, one inbox.

  • 500 conversations a month
  • 1 site
  • Web widget and email
  • Human handoff included
  • Knowledge base import
Book a demo

Scale

from €599 per month

Volume, control and a person to call.

  • 10,000 conversations a month
  • Unlimited sites
  • Per-tenant model choice
  • Custom connectors
  • SLA and onboarding
Book a demo

Over your allowance it is €0.08 a conversation, flat — no cliff, no renegotiation.

Bigger than Scale, or need it on your own infrastructure? Talk to us.

See it answer your customers' questions

Thirty minutes, your own help centre, live.

Book a demo

Questions people ask first

How long does setup take?
Import your help centre, paste one script tag into your site, and the widget is live. Connecting WooCommerce so Buddy can read real orders adds about ten minutes. Most of the work is deciding your escalation rules, not the installation itself.
What happens when Buddy cannot answer?
It escalates to a human, following the rules you write, and only when someone is actually online to take the conversation. If nobody is available it says so plainly and keeps helping with whatever else it can, rather than inventing an answer to fill the gap.
Which languages does it support?
Each tenant sets its own locale, so Buddy answers in that language and formats dates and numbers the way that market expects. A Danish store gets Danish replies and Danish date formats without you writing a second set of help articles.
How much does SupportBuddy cost?
Plans start at €49 a month for 500 conversations and run to €599 for 10,000, with a flat €0.08 per conversation beyond your allowance. You are billed per conversation rather than per resolution, and handing a conversation to a human is never charged as an extra.
What happens to our customers' data?
Conversations, articles and connector data are stored per tenant and are never shared between customers. Every message the bot handles is screened for prompt injection before it reaches the model, and every answer, tool call and token is logged so you can audit exactly what happened.