AIWebsite Chat Agentby Help Support Desk
Source-backed AI chat for support-ready websites

Launch a website assistant you can prove before customers use it.

Turn approved pages, FAQs, files, sitemaps, and supported product feeds into no-send testable answers with owner-visible evidence, review states, and safe handoff boundaries.

Source-backedanswers from approved knowledge, not broad web guesses
No-send testingpreview the same answer engine before launch
Review statesweak, risky, or unsupported answers become actions
Start free

No credit card needed to start. Paid plans upgrade by card only when approved and configured.

Review build published

Quick-launch gap work is live for inspection.

This page now surfaces the product position agreed in the competitor/spec work: safe website-trained AI chat, owner-approved knowledge, no-send testing, visible limits, and gated billing.

Open review portal

What changed

Launch cockpit, current-plan billing state, no-charge Free activation, transcript controls, widget/readback persistence, and go-live checklist feedback are published for review.

What is still gated

Card payments, customer sends, public indexing/campaigns, and final launch approval remain blocked until the review pass is complete.

What to review next

Run one real account through knowledge import, widget install/detection, safe test chat, conversation review, and final checklist evidence.

Every answer has a review trail.

AI Website Chat Agent connects the visitor question, grounded answer, deterministic evidence state, missing-knowledge action, and safe widget test in one reviewable workflow.

Illustrative visitor experience
Example visitor asks

“Which delivery options do you offer for garden furniture, and can I return it if it does not fit?”

Example source-backed answer

“When matching approved delivery and returns pages exist, the assistant can answer from those sources and show the owner what it used before go-live.”

Returns policyDelivery FAQProduct feed
Owner viewDeterministic evidence state
Source manifestApproved sources, import state, and excluded/unsafe targets stay visible
Traceable
Transcript reviewVisitor question, answer, citations, and follow-up visible to the owner
Reviewed
Safe widget testPreview the exact widget response before go-live
No-send
Example improvement action: add a clearer “large-item collection” FAQ before launch.

Product surfaces you can actually inspect

01Dashboard

Launch cockpit

Website state, knowledge approval, widget setup, and next safe action in one owner view.

02Manifest

Source approval

Pages, files, sitemaps, supported feeds, import status, and refresh evidence stay visible before answers go live.

03Review loop

Gap insight

Unanswered, weak, risky, or off-brand answers become specific knowledge or instruction updates instead of vague chatbot logs.

Content resources

Latest resources

The content space is ready; placeholder links point to honest empty-state pages until reviewed posts are published.

View all resources

Your website has the answers. Customers need them at the right moment.

Reduce repeat support

Answer common questions from reviewed pages and product data rather than generic chatbot guesses.

Expose missing knowledge

Unanswered, weak, risky, or off-brand chats become concrete source, FAQ, and instruction actions for the site owner.

Launch safely

Preview the widget, inspect citations, and verify handoff boundaries before opening it to visitors.

From website knowledge to working assistant

Add your website

Create your account and connect the site the assistant should support.

Approve knowledge

Import pages, FAQs, files, sitemaps, and supported feeds through a reviewed source workflow.

Shape the widget

Match greeting, colours, prompts, privacy text, and lead capture to your brand.

Test safely

Ask realistic questions in no-send preview before go-live.

Improve continuously

Use conversation review to find missing or weak answers and tune instructions.

Import approved pages, FAQs, files, sitemaps, and supported product feeds

The assistant is built around source approval and visible ingestion state. Product catalogue counting and ingestion manifests keep larger catalogues measurable instead of pretending every feed is equal.

Grounded knowledge

Public pages, FAQ and policy content, uploaded files, XML sitemaps, and product feeds where supported.

Brand control

Configure assistant name, greeting, colours, prompt chips, privacy copy, and lead capture settings.

Human review loop

Review transcripts, leads, unanswered/weak answers, missing knowledge, and instruction suggestions.

Start free, prove the setup, then upgrade when usage grows.

Free activates with no credit card needed. Stripe Checkout only starts when the billing gate is approved, while higher tiers unlock more chats, knowledge, websites, team seats, refresh frequency, and catalogue capacity as product-count limits land.

Free

£0/mo
  • One website
  • Small approved knowledge set
  • No credit card needed
Start Free

Starter

£29/mo
  • One live website
  • More chats and knowledge
  • Card checkout when approved
Pay by card

Growth

£79/mo
  • Growing sites and teams
  • More websites and usage
  • Card checkout when approved
Pay by card

Pro

£199/mo
  • Larger catalogues
  • Multi-site/team usage
  • Priority workflows
Pay by card

Final catalogue/product limits follow the ingestion manifest work. No live charge is created until a paid card upgrade is explicitly approved and completed.

Built around approval, boundaries, and visibility.

Approved sources first

No broad public crawl is promised. Knowledge comes from scoped, reviewed sources.

Billing clarity

Free needs no credit card. Paid plans use card checkout only when the billing gate is approved.

No hype claims

No fake metrics, live-stock promises, unsupported compliance claims, or unsupported confidence percentages.

Questions before launch

What makes the assistant safer than a generic chatbot?

It answers from approved website knowledge, exposes source and evidence review state to the owner, and turns weak or unsupported answers into improvement actions.

Can I test before visitors see it?

Yes. Safe Test Chat uses the live answer engine in a no-send preview so you can inspect responses, citations, and review state before go-live.

Are paid plans live today?

Free starts with no credit card needed. Paid upgrades stay behind approval gates until card payments are explicitly configured and approved.

Ask the AI assistant before you commit.

Use our own no-send AI assistant for sales questions, onboarding guidance, help, support, testing, knowledge import, and go-live boundaries without triggering customer messages, live billing, DNS, ads, or deployments.

Start free