v1.4 dropped · MIT · self-hosted

I LIKE BIG CHATS
AND I CANNOT LIE.

Open-source AI chat for your site. Grounded in your own published content — never freestyling your pricing — with inline forms, buttons, and confirms when text alone won't cut it.

Get the embed →
Sir Chats-a-Lot
Sir Chats-a-Lot
online · grounded in this site
Yo. I'm Sir Chats-a-Lot — ask me anything about the widget.
Try one
01Surfaces

Beyond text.
Forms. Buttons.
Confirms.

Sir doesn't just talk back. When a sentence isn't enough, he renders the UI the visitor needs — inline, in-thread, no detours to another page.

01 · BUTTONS
Action buttons
Got 2–5 paths the visitor might want? Sir renders them as tappable chips inline.
How do I install it?
Two paths — pick one:
NPM packageScript tagDocker
02 · FORM
Inline form
Need an email, a name, a quick brief? Sir generates the form and ships the submission to your webhook.
Can I get a demo?
Drop your email and I'll send it.
Get the snippet
name@yoursite.com
Submit →
03 · CONFIRM
Confirm dialog
Destructive or irreversible action coming? Sir asks first — before doing anything you'd regret.
Clear all my messages.
Heads up — this wipes the thread.
⚠ Confirm
Clear all messages?
Yes, do itCancel
02How it works

Three steps.
Five minutes.

01
Drop in the embed
One script tag, or one npm install. Five-minute setup, zero infrastructure.
<script src="sir-chats-a-lot.js" />
02
Point Sir at your docs
Connect your published content through the Gnosys knowledge layer. Sir reads only what you publish.
sir.gnosys = "/docs/**/*.md"
03
Visitor asks, Sir answers
Grounded answers and dynamic surfaces — forms, buttons, confirms — rendered in-thread.
// no hallucinated pricing.
03Install

Drop it in.
Done in five.

app/scal.tsx
# 1. install the package
npm install sir-chats-a-lot

# 2. mount it in your app (Next.js App Router)
import {
  SirChatsALotProvider,
  SirChatsALotLauncher,
  SirChatsALotWidget,
} from 'sir-chats-a-lot';
import 'sir-chats-a-lot/styles';

<SirChatsALotProvider productName="My SaaS" api={api}>
  <SirChatsALotLauncher />
  <SirChatsALotWidget />
</SirChatsALotProvider>
What to bring
  • ·An LLM API key (Anthropic, OpenAI, etc.)
  • ·Your published content (markdown, HTML, sitemap.xml)
What we ship
  • ·The widget (React)
  • ·A `gnosys web` CLI to index your content
What we don't
  • ·Hosted SaaS — you own the runtime
  • ·A database — knowledge is a static JSON index
04Features

Built like a brick.

01
Grounded answers
Pulls from your published content via the Gnosys knowledge layer. No bot inventing your pricing.
02
In-chat surfaces
Forms, buttons, confirm dialogs — rendered inline when text alone won’t cut it.
03
Self-hosted
BYO LLM key. Your data never touches our servers because we don’t have any.
04
MIT open source
Audit it. Fork it. Ship it. No vendor lock-in, no per-seat tax.
05
Drop-in embed
One script tag, or one npm package. React, Vue, plain HTML — anywhere.
06
Themeable
Token-driven theming. Match Sir to your site without writing a line of CSS.
05Use cases

Wherever
you got chat.

Sir is built for sites where the visitor's question deserves a real answer — and the answer occasionally needs a button, a form, or an "are you sure?" attached to it.

SaaS docs
Answer support, not tickets
Sir handles the "how do I…" stack out of your support inbox. Grounded in your docs. Surfaces a contact form when he genuinely doesn’t know — instead of guessing.
See case study →
E-commerce
Pre-purchase concierge
Sizing, shipping, returns — answered from your published policies. Renders product-picker buttons. Confirms with the visitor before doing anything irreversible.
See case study →
Agency / portfolio
Brief in three messages
Visitor describes a project; Sir surfaces a scoped intake form. Submission lands in your inbox while they’re still warm — no contact-page detour.
See case study →
06Changelog

Always shipping.

Full changelog →
v1.4.0
May 8, 2026
Surfaces 2.0
New confirm-dialog surface. Forms now support multi-step. Bundle shrunk by 18%.
v1.3.2
Apr 21, 2026
Streaming markdown
Code blocks and tables now render as Sir streams — no more 'wait until the end' jank.
v1.3.0
Mar 30, 2026
Gnosys integration
First-class adapter for the Gnosys knowledge layer. Ground answers in published content with a single config line.