Skip to main content

Custom chatbot development company for a website bot that answers from your content

Boffin Coders is a custom chatbot development company that builds a website chatbot on your own knowledge base: your pages, help articles and price rules. It answers from that content, links the page it used, and says when a question needs a person. The assistant in the corner of this page is built the same way, so you can test it before you call us.

A chatbot build starts at $4,000 · $25-40 an hour for shorter work · a developer by the month from $4,000 a month
  • Answers from your own content
  • Links the page it used
  • Website first, other channels on request

Running a clinic, a trade or a shop and need a website that brings enquiries? That is our local business line.

Website assistant

Answers from your own pages

Can your chatbot answer from our help pages?
Yes. It searches your pages for each question and links the one it used.From: Chatbots page
And on WhatsApp?
Yes. The same knowledge base connects to WhatsApp, with the same rules and the same handover to a person.

Illustration of a conversation. The real assistant is in the corner of this page.

Proof first

The chatbot on this page is the demo

No recording and no mock-up. The assistant is running in the corner of this page, answering from our own content, and you can ask it anything.

The live example

The assistant on this website

The gold button in the bottom corner opens an assistant that answers from this website’s own content. Ask it what a website costs, whether we work with agencies, or how we would build a chatbot for you. It answers from the same pages you can read, and links them.

It is also where we learnt the hard part. It used to send the model the first 23% of our content and nothing else, so a rule near the end of the file never reached it. It now picks the sections that match each question and always sends the ten that hold our prices and rules.

Or use the gold button in the bottom right corner of any page.

Knowledge base
This site’s pages, in 216 sections
Search
BM25 keyword scoring on our server
Always sent
10 pinned sections with prices and rules
Model
A hosted model, called through an API
Channel
This website only

Our pages are public, so a hosted model is fine for us. For private content, the model can run on your own server.

Where we are

What is running, and how we build the rest

The stage of our chatbot work, stated plainly, with the channels we connect on request and the principles every bot we build keeps.

Running for ourselves

This website’s assistant

Live on every page of this site, answering from our own content with pinned rules that are never dropped. It is a chatbot you can put your own hardest questions to, today.

On request

Other channels

WhatsApp, Slack, Microsoft Teams and SMS can all be connected to the same knowledge base, through each platform’s own API, with the same pinned rules and the same handover to a person. The answers stay the same whichever channel the question arrives on.

Our principle

An assistant that says what it is

Refunds and account changes stay behind a confirmation step or a person, so a misread message cannot change anything on its own. The bot never passes for a person: it says it is an assistant, answers from your own content, and hands over when a question needs one.

What we build

What a custom chatbot development company should build for you

Six things we build into yours. The assistant on this site does each of them today, on web pages rather than PDFs, so nothing on the list is a promise we cannot show you.

Answers from your content

Your pages, help articles, PDFs and price list, split into sections and searched for every question. It answers from what it found and nothing else.

Links the page it used

Every answer points at the page it came from, so a visitor can read the full version and your team can see why the bot said what it said.

Rules it never drops

Prices, promises you do not make and topics it must not touch go into every request, outside the search, so they are there on the question that needs them.

Hands over to a person

When a question needs a quote, a judgement or an exception, it says so and offers a call or an email to a person instead of guessing.

Stays out of the way on a phone

On a small screen the launcher waits until the visitor has scrolled past most of the first screen, so it never covers the buttons at the top of your page.

Asks for details only after it helps

A visitor gets real answers first. Name and email are asked for later, when the conversation has turned into something worth following up.

How it works

A website chatbot with your own knowledge base, in four steps

Owners usually ask us for “a website chatbot with our own knowledge base”. These four steps are what that takes, and most of the time goes on the content, not the model.

Gather the content

Your pages, help articles, PDFs and price list, plus the questions your team answers by email every week. Those questions become the test the bot has to pass.

  • Content gathered
  • Real questions collected
  • Gaps in the content listed

Split it and pin the rules

The content is cut into sections the search can score. Prices, prohibitions and handover rules are pinned so they are sent with every question, whatever the search picks.

  • Sections indexed
  • Rules pinned
  • Tone written down

Test it on real questions

The bot answers the questions from step one, and we read every answer with you. Wrong ones are traced to a missing page, a missing rule or a search miss, and fixed at the source.

  • Answers read with you
  • Each failure traced
  • A pass rate you can see

Put it on the site

The widget goes on your site, placed so it never covers a button, with the handover to a person wired to your inbox or booking link. The engineers who build it can also join your team by the month.

  • Widget live
  • Handover wired up
  • Content updates made easy
Channels

Where it runs, stated plainly

Channels

Website first, the rest on request

The chatbot you can try today runs on this very website. The knowledge base behind it does not care where the question comes from, so the same bot can be connected to WhatsApp, Slack, Microsoft Teams or SMS. Each channel is added through the platform’s own API, with the same pinned rules and a handover that reaches your team where they already work.

For most small businesses the website is the right first channel anyway. It is where the questions arrive, it needs no approval from a messaging platform, and you can see every conversation in one place.

Website widget
Shipped, live on this site
WhatsApp
On request, same knowledge base
Slack or Microsoft Teams
On request, same knowledge base
SMS
On request, same knowledge base
What it costs

Published, not quoted on the call

In USD. Hourly and monthly rates move with the stack and the seniority, from a junior on routine work to a senior on complex builds. Nothing is quoted outside them, and if the number does not work for you, you have saved yourself a meeting.

A chatbot build starts at
$4,000
Scoped once, with the lines shown.
Short pieces of work
$25-40 an hour
By seniority and the work.
A developer by the month
From $4,000 a month
Month to month, a month’s notice either way.
Agency sprint
$2,000 per two-week sprint
Under your brand, invoiced after you see the work.

Every price on one page →

FAQ

Chatbot questions before you book

What it answers from, what it does when it does not know, where it runs, how long it takes and what it costs.

Yes, and that is the only kind we build. Your pages, help articles, PDFs and price list become the knowledge base, and the bot answers from them and links the page it used. It does not answer from the open internet. When your content changes, the knowledge base is rebuilt from it, so the bot never runs on last month’s prices.

Not answered here?

Not quite this job?

A chatbot is one of four AI jobs

If staff need answers from internal documents that must stay on your server, that is RAG and private LLM deployment.

If the questions arrive as phone calls or recordings, start with voice AI on self-hosted speech to text.

Not sure which? Our AI automation page sorts the four jobs on one screen.

How we actually build it

Why your RAG chatbot ignores the rules you gave it

Grounding is not a switch. When a knowledge base outgrows the context budget, search starts deciding what the model sees, and the rules you wrote are competing for space with everything else. This is how we measure it on the assistant on this site.

Read the write-up

Check it yourself

context budget ÷ corpus size

Ask the assistant first, then book a call

Put your hardest customer question to the assistant on this page. If you like how it answers, a 20-minute call is enough to scope one for your site, with the price.