Parelay
- CATEGORYLive chat · AI · Video support
- HOSTINGSelf-hosted, single Node origin
- AIBring-your-own LLM, per site
- BEST FORHealthcare, finance, regulated
White-label live chat, AI, and video support you host yourself — chat, AI, recordings and all. Nothing in the conversation path is a vendor you don’t control. Built for healthcare, finance, and anywhere data can’t leave your walls.
The support stack you actually own.
The same conversations regulated teams can’t hand to a SaaS — chat, AI, voice, video, recordings — running entirely on infrastructure you control. It pairs naturally with our AI agents and healthcare engineering practices.
Your data never leaves
Chat, AI, video, and recordings all stay on your own servers.
No vendor in the path
Zero third parties between you and your customers.
One backend to run
A single origin serves the app, API, widget, and media.
Compliance-ready by design
Built for healthcare, finance, and regulated work.
The agent desk and the widget, side by side.
The dashboard your team lives in, the visitor widget on the page, and the admin screens that white-label both — all running over one backend you host.
Built for every side of the conversation.
Visitors get instant help, agents get one screen for everything, and admins configure it all without touching host-page code.
For your visitors
- Answers in seconds
The AI replies from the very first message — no waiting for someone to come online.
- A real human when it matters
An agent takes over mid-thread with no banner, no reintroduction, no lost context.
- Voice or video on the page
One tap accepts a call — no app to install, no link to click.
- See the agent’s screen, live
Screenshare arrives as its own tile next to the camera feeds, in real time.
- Picks up where you left off
Returning visitors are recognised and the thread survives a page refresh.
- Files and emoji, naturally
Drag a file in or drop an emoji from the same composer the agent uses.
- Floating call window
Keep a call up in picture-in-picture while you keep browsing the page.
- Asked once, never again
A single quiet consent gate before the first video call — not a wall of legal text.
For your agents
- Every visitor, every site, live
Who’s on, where, for how long, and what they’ve been doing — all at a glance.
- Take over from the AI in one click
Soft-locks keep two agents from claiming the same visitor at once.
- Many chats from one screen
The active rail surfaces unread, ringing, and typing signals so nothing slips.
- AI summaries on your terms
Fold in the chat, your notes, the audio transcript, and free-text additions.
- Video, screenshare, recording — one stage
Three layouts plus a live recording indicator; each call is its own playable file.
- A timeline of every page viewed
The visitor’s full journey this session, each step a clickable link.
- Know the visitor before you reply
Browser, OS, screen, language, timezone, IP, and location, captured on first sight.
- Capacity-aware controls
Per-row start buttons disable, with reasons, when you’re at your video or chat cap.
- Install as a desktop app
A PWA that survives tab close, with desktop push for new arrivals and chats.
For your admins
- Sites and agents in one window
Manage everything from the same shell your team takes calls in.
- White-label, per site
Brand colours, fonts, corners, launcher, and copy — your widget on your domain.
- Bring your own LLM
Provider, model, and system prompt are configurable, including a self-hosted gateway.
- Push events to your CRM
Webhooks on visitor.identified and session.ended, plus a branded chat-end email.
- Canned replies for the team
Author slash-command snippets once; every agent sends them instantly.
- Pre-chat capture, your way
Classic or conversational, per field, with server-side phone + email validation.
- Embed token + origin allowlist
Control exactly which domains can load your widget.
An AI that answers — and a human one click away.
The AI replies first and hands off cleanly to a person, with WebRTC voice, video, and screenshare built into the page. Bring your own model — see our AI/ML engineering bench if you want one stood up.
- 01 · FEATURE
AI-first by default
Visitors get instant answers; the widget clearly labels that they’re chatting with an AI assistant.
- 02 · FEATURE
Provider-agnostic
Anthropic, OpenAI, or an in-house gateway — chosen per site, with no external AI required.
- 03 · FEATURE
Seamless human handoff
A human can take over at any point; visitors can ask to escalate without an abrupt interruption.
- 04 · FEATURE
Post-session summaries
AI summaries that fold in chat, agent notes, the audio transcript, and free text — saved to the conversation.
- 05 · FEATURE
In-page WebRTC voice & video
No install and no third-party meeting link; screenshare renders as its own real-time tile.
- 06 · FEATURE
Recording per call
Each call recorded as its own file with thumbnails, durations, and in-browser playback.
White-label down to the pixel.
Every site is themed independently from its own config — set once in the admin, no rebuild and no host-page code. Your brand, your domain, your widget.
Primary, accent, secondary, border, and paper colours. First-class light and dark themes — not inverted.
Any Google font family; corner style (square, rounded, or pill) applied across the whole widget.
Bubble position in any corner; icon, text, or both; custom text and icon — even a looping launcher video.
Header, footer, greeting, and offline message — written in your voice.
Pre-chat style plus per-field collect/require for name, email, and phone, server-side validated.
Provider, model, and system prompt set per site — or no external AI at all.
Per-site embed token and an allowed-origins allowlist.
One backend. Nothing leaves it.
A single Node origin serves the agent app, the API, the real-time bus, the widget bundle, and media. Embed it with one script tag — the widget mounts in a Shadow DOM so it never collides with your styles.
- Self-hosted, single origin
Runs behind your own reverse proxy — the agent app, API, real-time bus, widget, and media from one Node origin.
- Zero third parties in the path
The only outbound calls are ones you configure: your CRM webhook, your email provider, your chosen LLM.
- You own retention & access
The controls regulated industries need — HIPAA, financial, and data-residency constraints.
- Quiet, single consent gate
One consent prompt before the first video call rather than a blocking popup.
<script
src="https://support.yourdomain.com/parelay.js"
data-site-token="st_your_site_token"
async defer></script>The bundle is ~15 KB gzipped and auto-captures SPA navigation for React, Angular, Vue, Svelte, and Next.js. Programmatic control via window.Parelay.open(), close(), and page().
Try the real thing, on its own site.
Parelay runs its own live deployment — open it and start a chat to see the AI answer and the handoff in action.
The questions regulated teams ask first.
Still weighing a build-vs-buy? Talk to us about a rapid POC or browse the rest of the product portfolio.
Is it really zero third parties?
Chat, AI, video, and recordings all run on your own backend. The only outbound calls are ones you configure — your CRM webhook, your email provider, your chosen LLM.
Can we use our own LLM?
Yes. The AI provider, model, and endpoint are configurable per site, including a self-hosted or in-house gateway. No external AI vendor is required in the path.
Is it self-hosted?
Yes. A single Node origin runs behind your reverse proxy and serves everything — the agent app, the API, the Socket.IO bus, the widget bundle, and media.
Is it suitable for healthcare or finance?
Conversation data stays on your infrastructure, and you control retention, access, and the AI backend — which is what regulated industries need.
Is it white-label?
Fully. Your brand, domain, colours, fonts, and copy. Customers never see a vendor name in front of the widget.
Own the conversation, end to end.
Tell us where your data has to live, and we'll stand up Parelay under your brand — chat, AI, voice, video, and recordings, all on infrastructure you control.