Engage platform

Your website. Your knowledge. One useful conversation.

Engage connects the information a business already publishes with an AI agent its customers can talk to. It answers from your pages, shows which page it used, and refuses when your content does not cover the question.

The system

The parts that exist, and a layer that does not yet.

The spine below is built and running today, and so is what happens when a visitor asks to be contacted. The layer beneath them is where Engage is going, and it is drawn differently because it is not here yet.

The conversation system

Available now

Running today, end to end, on any public website.

  1. WebsiteA bounded crawl of your own domain
  2. KnowledgePages extracted, split and embedded
  3. AI agentRetrieval, then a grounded answer
  4. ConversationThe exchange, with its citations

When a visitor wants to be contacted

Available now

Running today, through the widget on your own website.

  1. RequestThe visitor asks to buy, book or be contacted, or presses Contact us
  2. LeadTheir details and message, sent with consent
  3. EmailTo your owners and admins with a verified address

Expanding into

Not yet built

None of this is built. Nothing scores a visitor or a lead, and there is no routing, booking or integration in the product today.

  1. IntentScoring a commercial signal
  2. LeadRouted to the right person, or into another system
  3. Business actionBooking, follow-up
Website ingestion

It starts with an address, not a configuration form.

You give Engage a URL. It reads what is publicly reachable on that domain and nothing else.

  1. Discovery

    Links are followed within your own domain, behind a policy that resolves DNS and refuses private addresses.

    Same-origin only
  2. Fetch

    Pages are retrieved with a bounded page limit, so a large site does not become an unbounded crawl.

    Bounded
  3. Extract

    The DOM is parsed and the readable content separated from navigation and boilerplate.

    Parsed, not scraped
  4. Persist

    Clean content is stored with the URL it came from, which is what later lets an answer cite a page.

    Source retained
Before you ask

The questions we get first.

Is this just a chatbot on my website?

A generic chatbot answers from a model's general knowledge and will happily invent your opening hours. Engage retrieves passages from pages you published, answers from those, and names the page it used. When your content does not cover the question it says so rather than guessing.

Do you train a model on my website?

No. Nothing is trained or fine-tuned. Your pages are split into passages and stored as embeddings so the relevant ones can be found at question time and supplied to the model as context. Removing your content removes its effect immediately.

What does it do when it does not know?

It returns an explicit refusal with no citations, rather than a confident paragraph. That behaviour is part of the API contract: a supported answer always carries at least one citation, and an unsupported one always carries none.

Which pages does it read?

Publicly reachable pages on the domain you submit, up to a page limit, following links within that same origin. The URL policy resolves DNS and refuses private address ranges, so it cannot be pointed at an internal network.

Can I put it on my own site today?

Yes. With an agent in a workspace you own, turn on installation in the console, list the website addresses allowed to show it, and paste the snippet it gives you into your pages. The widget loads only on the addresses you list.

The console

What you can open today, and what is still a name.

Overview

Your workspace at a glance: agents, setup progress, and what each area is waiting on.

Agents

Create an agent, open it, rename it, publish, pause or archive it.

Settings

Members, roles and invitations for the workspace.

Usage

What your workspace has consumed, and what it cost. Every model call lands in an append-only ledger and is priced against a versioned rate catalogue.

Leads

Every visitor who left their details, with their message and a status your team moves from new to contacted to closed.

Not yet available — the console area is not built: Knowledge

Knowledge

Crawling and indexing run today, but the console has no page for reviewing them yet.

Nothing is counted here, because there is nothing to count yet.

Not yet available — the console area is not built: Conversations

Conversations

Conversations on your website are kept for your workspace, but the console has no page for reading them yet.

Nothing is counted here, because there is nothing to count yet.

Reliability

Built to answer — and to refuse.

Useful AI doesn't need to pretend it knows everything.

Supported answer

Same-day service is available in Miami for selected areas.

Cited to Services → Emergency repairs.

Insufficient information

I don't have enough information to confirm that.

No evidence found. The agent does not guess to fill the gap.

Human escalation

I can have someone from the team help you with that.

The question and its context stay attached to the conversation.

Reliable beats confident-sounding.

Honest roadmap

What exists, what half exists, and what does not.

Three states, because two would be a lie in one direction or the other. This table is generated from the same audit the engineering team works against.

Real today
  • Website crawl and indexingAvailableBounded, same-origin, behind an SSRF gate
  • Grounded answers with citationsAvailableCitations are contractual, not decorative
  • Refusal when evidence is missingAvailableA refusal is a legitimate outcome, not an error
  • Accounts, workspaces and membersAvailableSessions, roles and invitations
  • Durable agentsAvailableCreate, publish, pause, archive; a demo cannot name one
  • Asking your own agent a questionPartly availableWorks in a public demo and in your widget; no console route yet
  • Conversation historyPartly availableKept for your workspace; no console page to read it yet
  • Usage and costAvailableRecorded, priced, and visible to your own workspace
  • Embeddable website widgetAvailableOne snippet, shown only on the addresses you allow
  • Lead captureAvailableEmailed to your verified owners and admins, listed in the console
Not built
  • Intent qualificationNot built yetNothing scores, ranks or routes a visitor or a lead
  • CRM, booking and workflowsNot built yetNo CRM sync, webhook, export or booking integration, and no provider chosen

Each of these needs work that has not started, and several need a provider we have not selected. None of them can be switched on for you today.

Next step

See it read a website you know.

The fastest way to judge this is to point it at your own site and ask it something only your pages could answer.

Build your agent