A customer asks, anywhere
A message arrives from your web widget, a QR code, or the phone. Session scope — which account, which location, which customer — rides along automatically.
Your business, answered.
Zynlab is the multi-tenant agent backend any business plugs into. It grounds every reply in your knowledge, calls your systems to actually get things done, and talks to your customers on the web, in a widget, and on the phone.
Drop-in compatible with the OpenAI /v1/chat/completions API — point your
existing client at Zynlab, no rewrite.
AI Assistant
Web & Voice
Can I book a room this Friday to Sunday? Two adults. 🙂
Done — I've reserved a King Suite for two, Fri–Sun. Your confirmation is #NK-4821
and I've emailed the details. Anything else for your stay?
Booked
King Suite
Fri–Sun · 2 nights
2 guests
conf #NK-4821
Knowledge-grounded
Calls your APIs
Voice with barge-in
One-script widget
Multi-tenant
Human escalation
Audit trail
From a message to a done thing.
Grounds & decides
The engine retrieves from your tenant's knowledge base and reasons over a curated, tenant-scoped toolset — never inventing what it wasn't given.
Acts on your systems
Declarative HTTP tools call your booking, CRM, ticketing, ATS, or PMS — dispatching a structured action with an idempotency key and an audit trail.
Confirms & escalates
The customer gets one clear sentence back. Complaints, edge cases, and anything unclear are handed to a human — never silently auto-actioned.
Answering guests — and booking them in.
NiksenPMS, a property-management platform, runs Zynlab as its guest-concierge engine — live on client properties' public pages, where it answers from each property's knowledge and books guests in directly.
- In production inside NiksenPMS — on client property pages
- Answers guests on rooms, rates, amenities, hours & policies
- Reserves, moves, and cancels bookings — right in the chat
- Grounded — it stays within what each property has published
Guest booking is live today; staff-side ops agents that act inside NiksenPMS are rolling out next. Read the case study · See NiksenPMS ↗
AI Assistant
NiksenPMS · Guest concierge
Any rooms for 2 next weekend? What's the rate?
A Garden Suite is open at ₱6,800/night, breakfast included. Shall I hold it for
Sat–Sun?
Yes, book it please.
Booked — confirmation #NK-5093, emailed to you. See you this weekend! 🌿
Booked
Garden Suite
Sat–Sun · 2 nights
₱13,600
conf #NK-5093
One engine. Any industry.
The concierge
Room service, housekeeping, late checkout, amenities, local tips — dispatched to the right team, 24/7.
The front desk
Appointments, hours, departments, directions — and a firm hand-off to clinicians for anything medical.
The hiring assistant
Open roles, requirements, application status, and interview scheduling for every candidate, tirelessly.
Whatever you run
Retail, SaaS support, services, internal help desks. Set the persona, point it at your knowledge, wire your tools.
The same assistant, four different hats — switch between them live in the demo.
Production depth, not a chatbot demo.
Knowledge grounding
Per-tenant RAG over your policies, catalog, hours, and docs. Answers cite what you fed it — and stay silent on what you didn't.
Action tools — your APIs
Declarative HTTP executors turn intent into real work in your systems. Onboard a new capability with config, not a code release.
One-script widget
Drop one script tag into your site for a white-label chat & voice widget — keyed by a browser-safe publishable key and an origin allowlist.
Voice with barge-in
A streaming speech pipeline takes calls in a natural, interruptible conversation — the same persona and guardrails as chat.
Conversation records
Opt-in transcripts and call logs with PII redaction on by default, retention you control, and a Messenger-style thread for your staff.
Multi-tenant by design
You own the integrating customer; they pass per-session scope. One engine core, clean isolation, per-key cost tagging.
Set it up yourself — no code
Upload your knowledge (PDF, Word, Markdown), shape the persona, wire tools and voice, and test the assistant live — all from the console. Going live, or changing behavior, takes no engineering.
Analytics that close the loop
Requests and token spend over time, your deflection rate, and — uniquely — the questions the assistant couldn't answer, so you know exactly what knowledge to add next.
Live web search
When the answer isn't in your knowledge base, the assistant can pull current information from the web — gated per plan, so it never guesses in the dark.
Voice
It can answer the phone.
The same engine, fronted by a streaming STT → reason → TTS pipeline. Customers call and get a warm, fluent answer — and they can talk right over it.
- Rings, picks up, and speaks — like a real phone line
- Barge-in: speak the moment you want, it stops and listens
- Checks in on silence and ends dead-air calls gracefully
- Built on open-source voice models (Whisper + Kokoro) — no per-minute STT/TTS fees
If you can call OpenAI, you can call Zynlab.
POST /v1/chat/completions
200 OK · streamed
curl https://api.zynlab.dev/v1/chat/completions \
-H "Authorization: Bearer zl_live_…" \
-H "Content-Type: application/json" \
-d '{
"model": "zynlab",
"stream": true,
"session": { "account": "acme", "location": "downtown" },
"messages": [
{ "role": "user", "content": "Can I move my appointment to Friday?" }
]
}'
# → streams a reply, calls your tool,
# → and returns a structured action + audit id.Bearer-auth per tenant. Same SSE shape as OpenAI — Open WebUI, LibreChat, Continue, and your existing chat UI work unchanged. Full docs →
Sized to your operation.
Start free
₱0 / month
~250 chats or ~20 voice calls a month — shared allowance, on the house
- Knowledge-grounded chat & voice assistant
- Live web search & guided playbooks
- Lead capture & human escalation
- Embeddable web widget, one script tag
- Operator console, settings & live metrics
Scale on usage
₱999 / month
+ ₱450 per 1M credits beyond your monthly allowance
- Everything in Free — no monthly cap
- Your brand: custom theme, logo & no "Powered by" badge
- Custom integrations: connect your CRM, booking & systems
- Up to 25 concurrent voice calls (vs 3 on Free)
- Multi-tenant provisioning & priority support
Introductory pricing — we're finalizing plans, so these may change. Free gives you the full assistant — chat, voice & web search — up to a monthly allowance; Pro makes it yours (your brand, no badge), connects it to your systems, and removes the cap. Need higher volume or custom concurrency? Talk to us.
Get started
Put an agent on every channel.
Bring your knowledge and your systems. We'll wire up the brain that talks to your customers — on chat, in the widget, and on the phone.