Production voice agents, built by chatting.
Zapim Labs turns a plain-language brief into a live, emotion-aware AI voice agent — fluent in English, Hindi, and natural Hinglish — running on the omnichannel platform behind a decade of global voice and SMS.
All-inclusive per-minute pricing. TRAI and DLT compliant. Built for global markets, India-first.
Global voice aggregation
Global SMS aggregationParent companies of Zapim Labs — aggregating enterprise voice and SMS worldwide for over a decade.
New lab.
Decade-old backbone.
Zapim Labs is the AI brand of Vesper Telecom and Icon Global — global aggregators of enterprise voice and SMS for over ten years. The carrier relationships, routes, and delivery discipline that moved billions of messages now power generative voice AI.
IVR trees frustrate. Scripted bots stall. Customers just want to talk.
Legacy phone automation forces people through menus and keywords. Zapim Labs agents hold real conversations: they listen for intent, handle interruptions, switch languages mid-sentence, and read how the customer feels — then act on it.
- Before
"Press 1 for English. Press 2 for Hindi. Sorry, I didn't get that. Main menu."
- After
"Order kal deliver ho jayega? — Yes, it's out for delivery tomorrow before 6 pm. I've sent the tracking link on WhatsApp."
- The difference
Resolution in one call, in the customer's own words, with a full transcript and disposition for your team.
AI Voice Agents — the flagship
Describe the agent. Zapim Labs builds it.
Write what you need in plain language — "a polite collections caller for two-wheeler loans in Hindi and English" — and get a production-ready agent with a voice, a policy, and guardrails. No flow-charts, no vendor stack to assemble.
- Hears
- Understands
- Feels
- Responds
Emotion-aware speech
Our proprietary text-to-speech doesn't just pronounce, it performs — warmth for service, firmness for collections, energy for sales.
Native Hinglish
Customers switch between Hindi and English mid-sentence. So do our agents, without missing intent or a beat.
Live sentiment
Customer emotion is analysed during the call and fed to the model with no added latency, so the agent adapts in real time.
Voice cloning
Launch with a signature brand voice, cloned on the fly with consent, consistent across every conversation.
Grounded on your knowledge
Retrieval-augmented answers come from your documents and policies — accurate, current, and on-brand.
Guardrails built in
Scoped behaviour, escalation to humans, and full call auditing keep every conversation inside policy.
Every conversation, one platform, one wallet.
Voice is one channel in a larger relationship. Zapim runs WhatsApp, RCS, SMS, Email, and AI Voice from a single console — shared contacts, shared templates, shared balance — so a call can hand off to a WhatsApp message and back without losing context.
- Unified wallet and billing across every channel, in INR or USD.
- Shared contacts and number pools so audiences stay consistent.
- Channel handoff — voice confirms, WhatsApp documents, SMS reminds.
- One analytics view of delivery, engagement, and outcomes.
Multi-tenant by design: run it as your brand, or let partners run it as theirs.
From idea to live calls in three steps.
Describe
Tell the builder what the agent does, who it calls, and how it should sound — in a chat, in plain language.
Ground
Connect your knowledge base, CRM, and policies. The agent answers from your truth, inside your guardrails.
Deploy
Go live on real phone numbers, at scale, with outcomes flowing to your CRM and follow-ups on WhatsApp, SMS, or email.
Solutions by industry
Tuned for the conversations your industry runs on.
Collections that stay courteous — and compliant.
Every EMI call made on time, in Hindi, English, or Hinglish, with a tone your compliance team approved and an audit trail on all of it.
- Reminders that convert — payment links land on WhatsApp before the call ends.
- Promises, captured — promise-to-pay written straight into your collections system.
- Soft buckets, automated — human callers freed for the hard cases.
Kill RTO before it ships.
A confirmation call minutes after checkout separates real orders from expensive mistakes — then keeps the buyer informed to the doorstep.
- COD confirmed in minutes — fakes cancelled before they cost you freight.
- Deliveries rescued — NDR and reattempt calls recover what would return.
- Journeys, not one-offs — voice confirms, WhatsApp tracks, SMS reminds.
Every appointment kept, with care.
Reminders, report notifications, and follow-ups handled warmly — and anything clinical handed straight to your staff by policy.
- No-shows, reduced — evening-before reminders with one-tap reschedule.
- Reports, communicated — collection details sent on WhatsApp.
- Boundaries, enforced — medical advice is off-limits by guardrail.
Call the lead while intent is warm.
An agent that dials within seconds of the enquiry, qualifies budget and locality, and books the site visit — before the lead goes cold.
- Instant response — every portal enquiry called back in seconds.
- Qualified, scored — budget, configuration, and timeline into your CRM.
- Visits that happen — confirmations plus day-before reminders.
Admissions counselling at scale.
Thousands of near-identical conversations, each handled patiently in the family's language — with fee reminders that stay respectful.
- Enquiries answered instantly — programmes, eligibility, and deadlines.
- Applications completed — incomplete forms nudged to submission.
- Fees followed up — links and confirmations on WhatsApp.
Sell it as your own.
Agencies, aggregators, operators, and system integrators run Zapim under their own brand — own domain, own pricing, own client tree — on wholesale rates from a team that has lived the aggregation business for a decade.
- White-label — your domain, your logo, your invoices.
- Reseller trees — manage clients and sub-resellers with isolated tenants.
- Wholesale economics — prepaid wallet, margins you control.
Good to know
Questions leaders ask before the first call.
What is an AI voice agent?
Which languages do your agents speak?
How is Zapim Labs priced?
Is my data safe, and are campaigns compliant in India?
Put an agent on your next call.
See a live conversation on your own use case — in English, Hindi, or Hinglish — and leave with pricing you can take to the board.
Voice agents that sound like your best rep.
Emotion-aware, interruption-friendly, fluent in English, Hindi, and natural Hinglish. Built from a plain-language brief, grounded on your knowledge, and governed by guardrails — then dialed at scale.
One all-inclusive per-minute rate. Outcomes pushed to your CRM. Compliance handled.
From prompt to production
You describe the conversation. We handle everything underneath.
Behind one chat box sits a complete calling operation: speech recognition tuned for India, a self-hosted language model, our proprietary emotion-capable voice, telephony, campaign management, and reporting — engineered together so nothing is left for you to assemble.
- Hears
- Understands
- Feels
- Responds
Emotion-capable voice
Our own text-to-speech performs the script: reassuring for support, confident for sales, appropriately firm for collections. Tone is a setting, not an accident.
Real-time sentiment
Customer emotion is analysed live and fed to the model with no added latency, so the agent softens, speeds up, or escalates while the call is still happening.
Hinglish, natively
Mid-sentence switches between Hindi and English are understood as one language, the way your customers actually speak.
On-the-fly voice cloning
Give your brand a signature voice — cloned with consent, consistent across millions of calls, exclusively yours.
Grounded answers
Retrieval-augmented generation draws on your documents, policies, and product data, so answers are accurate and current — not improvised.
Guardrails framework
Scoped topics, forbidden actions, human-escalation rules, and full call auditing keep every conversation inside policy — essential for regulated teams.
Natural interruptions
Barge-in handling means customers can talk over the agent and be heard, keeping conversations fast and human.
Channel handoff
The call confirms; WhatsApp or SMS documents. Links, locations, and receipts arrive before the customer hangs up.
Every call, analysed
Transcripts, dispositions, sentiment, and outcomes on 100% of calls — not the 2% a manual QA team can sample.
Engineered end to end
Latency is a feature. So we own the pipeline.
Every stage — hearing, thinking, speaking — runs on infrastructure we control, tuned as one system for sub-second responses that keep conversations natural.
Deep dive on the stack — our TTS, self-hosted models, and safety work — on the Technology page.
What teams run on it
One builder, every calling workflow.
Inbound support
Answer every call instantly, resolve routine queries, and route the rest to humans with full context.
Outbound sales
Qualify, pitch, and book meetings at a pace no team can match — with a tone that never tires.
Collections and reminders
Firm, respectful EMI and dues conversations that log promises to pay and send payment links on WhatsApp.
Lead qualification
Call new leads within seconds of the form submit, score them, and hand hot ones to sales.
Operations calls
Delivery confirmations, appointment reminders, renewals, and verifications — automated end to end.
Surveys and CSAT
Conversational feedback at scale, with sentiment analysed on every response.
Fits your systems
Outcomes land where your team already works.
Every call produces structured results — disposition, transcript, sentiment, promised actions — pushed automatically into your CRM or any system via webhooks and REST APIs.
Good to know
Before you put it on the phone.
How accurate are the agents with Hinglish and Indian accents?
Can customers interrupt the agent?
What happens when a call needs a human?
How do you handle recording and consent?
How long does it take to go live?
Hear it handle your hardest call.
Bring a real scenario — a tough collection, a confused customer, a Hinglish rant — and watch the agent hold the line.
Every conversation. One platform.
WhatsApp, RCS, SMS, Email, and AI Voice — run from a single console with shared contacts, shared templates, and one prepaid wallet. Built multi-tenant, compliant in India, and ready for global routes.
TRAI and DLT managed. WhatsApp Business API. Carrier-grade delivery, at home and abroad.
Channels
Five channels, one operating rhythm.
Each channel is a full product on its own — and better together, because audiences, templates, balances, and analytics are shared across all of them.
WhatsApp Business API
Templated campaigns, a shared team inbox, and rich media on the channel your customers open first. Business verification, display-name and template approvals handled with you; Meta conversation charges passed through transparently.
- Marketing, utility, and authentication templates
- Two-way inbox with assignment and history
- Campaigns with audience targeting and scheduling
RCS Business Messaging
Branded, verified messaging with rich cards, carousels, and suggested actions — native in the default Android messages app, in one of the world's fastest-growing RCS markets.
- Verified sender with brand logo and name
- Rich cards, carousels, and quick replies
- Automatic SMS fallback preserves reach
SMS
The channel that reaches everyone, run properly: DLT-registered headers and templates, promotional-window enforcement, DND scrubbing, and OTP-grade delivery for transactional traffic.
- Transactional, promotional, and OTP routes
- DLT compliance enforced by the platform
- Global destinations on aggregation-grade routes
Campaign and transactional email alongside your messaging — same contacts, same reporting — so the whole journey lives in one place.
- Campaign builder with templates
- Transactional sending via API
- Engagement analytics in the unified view
AI Voice
Zapim Labs voice agents are native to the platform — same wallet, same contacts, same numbers — so calls, messages, and follow-ups behave as one conversation.
- Agents built by chatting, live in minutes
- Call outcomes trigger WhatsApp or SMS follow-ups
- All-inclusive per-minute pricing
Unified analytics
Delivery, engagement, and outcomes across every channel in one honest view — filterable by channel, campaign, and time, exportable when you need it.
- Channel-aware dashboards
- Campaign-level drill-downs
- Exports and API access to your data
Operations
Shared everything, so nothing falls between channels.
- One prepaid wallet funds every channel, in INR or USD, with GST-compliant invoicing.
- Shared contact lists and segments keep audiences consistent from SMS to voice.
- Number pools manage sender identities across calling and messaging in one place.
- Template library spans channels, with approval status tracked where approvals apply.
- Roles and permissions for marketing, support, finance, and admin — each seeing what they should.
Channel handoff, built in
A voice agent confirms the order; WhatsApp delivers the invoice; SMS reminds on the day. Journeys move across channels with context intact — that is the point of running them on one platform.
Architecture
Multi-tenant to the core — run it as yours, or as theirs.
Zapim was built as a multi-tenant platform from day one. Enterprises get isolated workspaces and role-based control; partners get something rarer — the ability to run the whole platform under their own brand, with their own client tree and pricing.
- Isolated tenants
Each business, partner, and client operates in its own walled workspace with its own data, users, and wallet.
- Three-tier trees
Platform, partner, client — partners manage clients and margins without touching each other's worlds.
- Your brand throughout
White-label partners run their own domain, logo, and invoices on Zapim rails.
Delivery and trust
A decade of aggregation discipline, applied to every message.
Deliverability is earned through carrier relationships, clean routing, and compliance done properly. That is the business Vesper Telecom and Icon Global have run for over ten years — and the standard the platform enforces by default.
DLT enforced
Headers, templates, time windows, and DND handled by the platform, not by hope.
Global routes
International sending on aggregator-grade connections, configured per destination.
Secure by design
Encryption in transit and at rest, role-based access, audit logs. Details on the Security page.
Managed onboarding
Migration, verifications, and approvals run with you by a team that has done it for years.
Good to know
Practicalities, answered.
Do you handle DLT registration and templates for India?
How does WhatsApp Business API onboarding work?
Where is RCS available, and is it worth adding?
Can we migrate from our current provider without downtime?
Can you send internationally as well as within India?
Bring every channel under one roof.
See your own use case running across WhatsApp, RCS, SMS, email, and voice — with one wallet and one view of the truth.
A voice AI stack we own, end to end.
Most voice platforms are assemblies of other people's models, glued together and marked up. We took the harder road: our own speech synthesis, self-hosted language models on our own GPUs, and a pipeline engineered as one system — faster responses, better economics, data that stays put.
The pipeline
Five stages, one latency budget.
A natural conversation gives you about a second to hear, think, and speak. We spend that budget deliberately at every stage — because a pause that feels fine in a demo kills trust on a real collections call.
Inside the stack
What we built, and why it matters to you.
Proprietary text-to-speech
Our voice model was built in-house to do two things most engines can't: perform emotion — warmth, urgency, firmness — on demand, and do it at unit economics engineered below one rupee per minute at the model layer. That efficiency is why our all-inclusive pricing is honest instead of aspirational.
Speech recognition, tuned for India
We run a best-of-breed Indic recognition stack — including leading Indian ASR — with active engineering on the hardest problem in the market: Hinglish code-switching, accents, and Indian names and numbers, recognised as one natural language.
Self-hosted language models
The reasoning layer runs on open-weight models hosted on our own GPUs. No per-token toll to a third party, no conversation data leaving our infrastructure to train someone else's model, and latency we control down to the millisecond.
Retrieval-augmented grounding
Agents answer from your documents, policies, and product data through a retrieval architecture built for voice — fast enough to fetch mid-conversation, disciplined enough to say "let me connect you to a specialist" instead of guessing.
Real-time sentiment and emotion
We analyse how the customer sounds — not just what they say — during the call, and feed that signal to the model with no added latency. Frustration triggers de-escalation; hesitation triggers reassurance; anger triggers a human.
Guardrails framework
Every agent runs inside an explicit policy: allowed topics, forbidden claims, mandated disclosures, escalation triggers, and hardening against attempts to manipulate the agent off-script. Every call is transcribed and auditable, so compliance teams review evidence, not anecdotes.
Latency engineering
Streaming at every stage, models co-located on our GPUs, and continuous benchmarking of the full round trip — because we measure what customers feel: the pause before the agent speaks.
Voice cloning
A consented sample becomes a production brand voice, consistent across languages and campaigns. Cloning without documented consent is refused by policy.
The ownership dividend
Why owning the stack beats assembling one.
Platforms that stitch together third-party recognition, models, voices, and telephony pass you four invoices, four latency hops, and four data processors. Ownership collapses all of that.
- Honest pricing. One all-inclusive per-minute rate, because there's no one downstream to pay.
- Predictable latency. One team owns the whole round trip and its regressions.
- Data control. Conversations are processed on our infrastructure under your agreement — not scattered across vendors.
- Faster iteration. When a customer needs a new voice behaviour, we change the model, not a contract.
| Assembled stacks | Zapim Labs | |
|---|---|---|
| Vendors per call | Typically 3 to 4 | One |
| Billing | Component rates that stack up | All-inclusive per minute |
| Hinglish handling | Afterthought | First-class |
| Data flow | Across multiple processors | Our infrastructure, your agreement |
| Beyond voice | Voice only | WhatsApp, RCS, SMS, Email included |
Security and residency
Enterprise controls, from day one.
Encryption
In transit and at rest, across calls, transcripts, and messages.
India residency
Processing and storage in India for domestic deployments; regional options for global rollouts.
Access control
Role-based permissions, tenant isolation, and audit logs on sensitive actions.
Dedicated options
Dedicated GPU capacity and private deployments for regulated enterprise workloads.
Full posture — compliance, certifications, and AI safety — on the Security & trust page.
Kick the tyres with the people who built it.
Bring your latency bar, your language mix, and your security questionnaire. Our engineers enjoy the hard ones.
Built for the conversations your industry runs on.
Every sector has a handful of calls and messages that decide its economics — the EMI reminder, the COD confirmation, the admissions counselling call. Zapim Labs automates exactly those, in your customer's language, inside your compliance rules.
Banking & financial services
Banking & financial services — collections that stay courteous, and compliant.
Collections and reminders are relationship moments handled at brutal volume. Our agents make every EMI call on time, in Hindi, English, or Hinglish, with a tone your compliance team approved — and log the outcome of every single one.
- EMI and dues reminders with payment links delivered on WhatsApp before the call ends.
- Promise-to-pay capture written straight into your collections system, dated and auditable.
- KYC and document follow-ups that chase politely until the file is complete.
- Policy and renewal calls for insurance and lending, at renewal-season scale.
- Soft-bucket collections handled by AI so your human callers focus on hard cases.
Ecommerce & D2C
Ecommerce & D2C — kill RTO before it ships.
Cash-on-delivery returns and missed deliveries quietly eat margin. A confirmation call within minutes of the order — in the buyer's own language — separates real orders from expensive mistakes, then keeps the customer informed to the doorstep.
- COD confirmation calls minutes after checkout, cancelling fakes before they cost you freight.
- Delivery-day coordination confirming availability and address details that couriers miss.
- NDR and reattempt calls recovering deliveries that would otherwise return.
- Returns and refund status answered instantly instead of queuing in support.
- Win-back and reorder nudges by voice, with the offer landing on WhatsApp.
Healthcare
Healthcare — every appointment kept, every result communicated, with care.
No-shows waste clinical capacity, and follow-up calls are the first thing busy front desks drop. Agents handle the routine with warmth and precision, and hand anything clinical straight to your staff.
- Appointment reminders the evening before, with one-tap reschedule into open slots.
- Report-ready notifications with collection details sent on WhatsApp.
- Post-visit follow-ups checking in on schedules your protocols define.
- Camp and screening outreach at community scale, in local languages.
- Front-desk overflow so no patient call rings out at peak hours.
Real estate
Real estate — call the lead while the intent is still warm.
Property leads decay by the hour. An agent that calls within seconds of the enquiry, qualifies budget and locality, and books the site visit — then reminds the day before — changes conversion math without adding headcount.
- Instant lead response calling within seconds of every portal or ad enquiry.
- Qualification of budget, configuration, locality, and timeline, scored into your CRM.
- Site-visit scheduling with confirmations and day-before reminders.
- Project updates to interested buyers as construction milestones land.
- Broker and channel-partner coordination handled without your sales team dialling.
Education
Education — admissions counselling at scale, fee follow-ups without friction.
Admission seasons bring thousands of near-identical conversations, and fee reminders strain relationships when done badly. Agents counsel patiently in the family's language, and remind firmly but respectfully — every time.
- Admissions enquiry handling answering programme, eligibility, and deadline questions instantly.
- Application follow-ups nudging incomplete forms to submission.
- Fee reminders with payment links and confirmation on WhatsApp.
- Parent communication for attendance, events, and announcements by voice and message.
- Alumni and event outreach at volumes no phone bank matches.
Not on the list?
If you can describe the call, we can build the agent.
Logistics, travel, hospitality, government service delivery, utilities — the builder is general-purpose by design. Describe the conversation in plain language and see a working agent the same day.
- Start with one workflow
Pick the call that hurts most. Pilot it in days, measure honestly, then expand.
- Keep your systems
Outcomes flow into the CRM and tools your team already uses.
- Grow across channels
Add WhatsApp, RCS, SMS, and email journeys on the same platform when you're ready.
Start with the call that costs you most.
Tell us the workflow and the language mix. We'll demo a working agent on your scenario, with pricing to match.
Sell voice AI and omnichannel — under your brand.
Agencies, aggregators, operators, and system integrators run the full Zapim platform as their own product: your domain, your pricing, your client tree, on wholesale economics from a house that has lived the aggregation business for over a decade.
Multi-tier tenancy. Prepaid wholesale wallet. Margins you set. Support that stays behind the curtain.
Three ways in
Pick the model that fits your business.
Reseller
Sell under the Zapim Labs brand with wholesale rates and your own client tree. Fastest path to revenue: our brand equity, your relationships, your margin.
- Wholesale per-minute and per-message rates
- Client management console and sub-accounts
- Co-branded sales collateral
White-label
The whole platform as your product — your domain, your identity, your invoices. Your clients never see us, and never need to.
- Full brand replacement, end to end
- Your retail pricing, set per client
- Priority support behind your team
Referral
Introduce customers, earn on their usage, and let us run everything else. Zero operations, real recurring income.
- No commitment to start
- Transparent revenue share on referred usage
- Upgrade to reseller anytime
The economics
Wholesale in. Your price out. The spread is yours.
Partner economics are deliberately simple: a prepaid wholesale wallet funds your entire tree, you set retail pricing per client, and your console shows usage, cost, and margin in real time. No revenue-share puzzles, no month-end surprises.
- Wholesale rates across voice minutes and every messaging channel.
- Per-client pricing control — premium for some, strategic for others.
- Live margin visibility by client, channel, and campaign.
- One wallet, whole tree — fund once, allocate as you choose.
We've been on your side of the table.
Vesper Telecom and Icon Global built their business aggregating voice and SMS globally for 10+ years — buying wholesale, selling with a margin, and living on deliverability. This partner program is the one we always wanted as buyers: honest rates, real isolation between clients, and a platform we're proud to put our name behind — or yours.
Enablement
Everything you need to sell it well.
Fast onboarding
Partner workspace, branding, and first client live in days, guided by our team.
Sales kit
Decks, demo scripts, and case-study templates — re-brandable for white-label partners.
Technical backup
Priority support and solution engineering on your bigger pursuits.
Growth reviews
Quarterly pipeline and margin reviews with your partner manager.
Who partners with us
Built for businesses that own customer relationships.
Good to know
Partner questions, answered straight.
How deep does the white-label go?
How do billing and margins work?
Is there a minimum commitment to become a partner?
Who supports my clients?
Can my clients have their own sub-accounts and users?
Put your brand on the front of it.
Tell us about your clients and your market. We'll share wholesale rates and have your partner workspace ready this week.
One rate per minute. Everything included.
No model fees, no synthesis fees, no telephony surprises. Prepaid credits in INR or USD, and a rate that means what it says — because we own the stack it runs on.
Starter
Self-serve. Prove it on real calls.
- AI voice agents, built by chatting
- Up to 5 concurrent calls
- WhatsApp, RCS, SMS, and email pay-as-you-go
- Transcripts, sentiment, and analytics on every call
- Email support
Growth
For teams putting calling to work.
- Everything in Starter
- Up to 25 concurrent calls
- CRM connectors: Salesforce, Zoho, HubSpot, webhooks
- Campaign tools and priority routing
- Priority support with onboarding assistance
Enterprise
For scale, regulation, and white-label.
- Everything in Growth
- Concurrency engineered to your peak
- 99.9% uptime SLA and named success manager
- Dedicated GPUs, private deployment, India residency
- White-label and multi-tenant partner trees
Indicative launch pricing shown for planning; final rates are confirmed on your order form. Taxes additional. USD figures are approximate conversions.
Estimate your spend
Do the math before the meeting.
Voice only; messaging billed per message. Estimates for planning, not a quote.
Messaging rates
Channel pricing, without the fog.
Meta conversation charges passed through at cost, plus a clear platform fee. No hidden markup on Meta's rates.
SMS
Domestic DLT-registered routes from ₹0.15 per message on slab pricing; international destinations quoted per route.
RCS
Per-message pricing with automatic SMS fallback, quoted on your volumes as operator rates apply.
Volume-based sending bundled with the platform — ask for slabs on your traffic profile.
Compare plans
What changes as you grow.
| Capability | Starter | Growth | Enterprise |
|---|---|---|---|
| All-inclusive voice rate | from ₹6/min | from ₹5/min | volume-based |
| Concurrent calls | 5 | 25 | engineered to peak |
| Channels: WhatsApp, RCS, SMS, Email | Included | Included | Included |
| Transcripts, sentiment, analytics | All calls | All calls | All calls |
| CRM connectors and webhooks | Webhooks | Full connectors | Custom integrations |
| Voice cloning | On request | Included | Included |
| Uptime SLA | Best effort | Priority | 99.9% contractual |
| Dedicated GPUs / private deployment | – | – | Available |
| India data residency guarantee | Standard | Standard | Contractual |
| White-label and partner trees | – | – | Available |
| Support | Priority | Named success manager |
Good to know
The fine print, in plain language.
What exactly is included in the per-minute rate?
Do prepaid credits expire?
How does messaging pricing work?
What do enterprise plans change?
Are there setup fees or lock-ins?
Get a number you can defend in the boardroom.
Share your volumes and we'll return a clean, all-inclusive quote — with the trial credentials to verify it on real calls.
The AI lab of a global telecom house.
Zapim Labs is a fully owned brand of Vesper Telecom and Icon Global — global aggregators of enterprise voice and SMS for over a decade. We didn't arrive at communications through AI; we arrived at AI through communications.
Our story
Ten years of carrying other people's conversations. Then we started building our own.
For 10+ years, Vesper Telecom and Icon Global have aggregated voice and SMS across global routes — negotiating with carriers, defending deliverability, and running the unglamorous plumbing that enterprise communication depends on. That business teaches lessons no amount of model training can: latency is felt, compliance is survival, and margins live in the infrastructure you own.
Zapim Labs is those lessons, rebuilt as product. First came Zapim, our multi-tenant omnichannel CPaaS. Then came the lab itself — our own emotion-capable speech synthesis, self-hosted language models on our own GPUs, and generative voice agents anyone can build by chatting. Launched for global markets, with India as our first and deepest one.
- The last decade
Global voice and SMS aggregation at enterprise scale — carrier relationships, routes, and delivery discipline.
- Then
Zapim: a multi-tenant omnichannel CPaaS unifying WhatsApp, RCS, SMS, email, and voice with white-label partner trees.
- Now
Zapim Labs: proprietary voice AI — emotion-capable TTS, self-hosted models, and agents built from a plain-language brief.
- 2026
Global launch across both product lines. India-first depth, world-ready infrastructure.
What we believe
The principles the lab runs on.
Own the stack
Models, voices, and infrastructure we control — because economics, latency, and data promises all trace back to ownership.
Ship what's real
Demos are easy; production is the product. We measure ourselves on live calls, not lab clips.
Compliance by default
A decade in telecom taught us that trust is infrastructure. DLT, consent, and auditability are features, not paperwork.
Customer economics first
All-inclusive pricing at rates our owned stack makes honest — value that survives the CFO's spreadsheet.
Two product lines, one platform
What we operate.
Zapim Labs — AI Voice Agents
Generative, emotion-aware voice agents fluent in English, Hindi, and Hinglish, built by chatting and governed by guardrails. Explore voice agents
Zapim — Omnichannel CPaaS
WhatsApp, RCS, SMS, email, and voice on one multi-tenant platform, with unified wallets and white-label partner trees. Explore the platform
Leadership
Founder-led, telecom-raised.
A founder-led team spanning telecom operations, platform engineering, and applied AI research — people who have negotiated carrier routes and trained the models now answering the phone.
Leadership
The people behind the platform.
Photographs and biographies drop in here — the frames below are sized and styled, ready for your final images.
Careers
Help teach machines to hold a conversation.
We hire engineers, researchers, and operators who like hard, measurable problems: shaving milliseconds off a speech pipeline, making Hinglish feel native, and building a platform partners bet their brand on. If that sounds like your kind of work, write to us.
- Voice AI research — TTS, ASR, and conversational modelling.
- Platform engineering — multi-tenant systems at telecom scale.
- Solutions and partnerships — putting agents to work across industries and borders.
See what a decade of telecom sounds like.
Meet the team, hear the agents, and judge the platform the way we do — on a live call.
Put an agent on the line.
Tell us what you're trying to do. Within one business day you'll hear from a human — and, if you'd like, from one of our agents, live on your use case.
We read, properly
A solutions engineer reviews your use case — not a routing bot — and replies within one business day.
You hear it live
A demo on your scenario, in your language mix, including a call from an agent to your own phone if you want one.
You get real numbers
All-inclusive pricing on your volumes, a pilot plan, and trial credentials to verify everything yourself.
Reach us directly
- Sales and demos — hello@zapimlabs.com
- Partnerships — partners@zapimlabs.com
- Support — support@zapimlabs.com
Offices in India (Bangalore, Gurugram, Noida), the UAE, Hong Kong and London. A Vesper Telecom & Icon Global company.
Free trial
Prefer to see for yourself?
Start self-serve: trial credits on your own workspace, an agent built from your first prompt, and real calls the same day. Choose "Free trial" in the form and we'll send credentials with your welcome note.
Voice and messaging, a few calls deep.
A clean REST surface over the whole platform: create agents, place and receive calls, send on every channel, and stream outcomes back over webhooks. Sandbox first, production when you're ready.
REST and JSON everywhere. Webhooks for everything that happens. Full reference in your dashboard.
# Create an agent from a plain-language brief
curl -X POST https://api.zapimlabs.com/v1/agents \
-H "Authorization: Bearer $ZAPIM_KEY" \
-H "Content-Type: application/json" \
-d '{
"name": "cod-confirmer",
"brief": "Confirm COD orders in Hindi or English, offer address edits, send tracking on WhatsApp.",
"voice": "warm-female-in"
}'Quickstart
First call in three steps.
Get keys
Create a workspace, grab a sandbox key from the dashboard, and keep production keys for later — the surface is identical.
Create an agent
POST a plain-language brief to
/v1/agents. The response includes the agent id and a test number to hear it immediately.Place a call
POST to
/v1/callswith the agent id and destination. Watch the webhook stream as the conversation happens.
# Place an outbound call
curl -X POST https://api.zapimlabs.com/v1/calls \
-H "Authorization: Bearer $ZAPIM_KEY" \
-d '{ "agent_id": "agt_8x2k", "to": "+9198XXXXXX21", "context": { "order_id": "ORD-10422" } }'Surface area
Everything the console does, the API does.
Agents
Create, update, and version agents; manage voices, guardrail policies, and knowledge attachments.
Calls and campaigns
Place single calls or launch batch campaigns with pacing, retries, and calling-window rules.
Messaging
Send WhatsApp, RCS, SMS, and email through one consistent request shape, with template management included.
Contacts
Manage lists, segments, and consent flags shared across voice and messaging.
Wallet and usage
Query balances and per-channel usage; reconcile programmatically against finance systems.
Reporting
Pull transcripts, sentiment, dispositions, and delivery data — or receive them as they happen.
Webhooks
Your systems find out the moment it happens.
Call started, transcript ready, promise-to-pay captured, message delivered, message read — subscribe once and receive signed JSON events for the moments your workflows care about.
- Signed payloads so you can verify origin.
- Automatic retries with idempotency keys.
- Per-event subscriptions — take only what you use.
{
"event": "call.completed",
"call_id": "cal_7f31",
"agent_id": "agt_8x2k",
"duration_sec": 74,
"disposition": "confirmed",
"sentiment": "positive",
"transcript_url": "https://api.zapimlabs.com/v1/transcripts/tr_91ac"
}Sandbox and production
Identical APIs, separate keys and data. Break things safely, then flip the key.
Security
Bearer keys with rotation, IP allow-listing on request, and audit logs on key usage.
Full reference
Endpoint-by-endpoint documentation, schemas, and examples live in your dashboard from day one.
Sandbox keys in minutes.
Start a trial workspace, grab a key, and have an agent talking before your coffee cools.
Trust is infrastructure. We build it like ours depends on it — because it does.
A decade of carrying enterprise traffic taught Vesper Telecom and Icon Global a simple rule: platforms earn the next message by protecting the last one. Zapim Labs applies that rule to conversations, data, and AI behaviour alike.
Data protection
Your conversations stay yours.
Encryption everywhere
Calls, transcripts, messages, and stored data are encrypted in transit and at rest, with key management under strict access control.
Residency options
Indian deployments are processed and stored in India by default; enterprise contracts can make residency a guarantee, with regional options for global rollouts.
Owned infrastructure
Language models and speech synthesis run on our own GPUs — conversation data isn't shipped to third-party model vendors as a side effect of every call.
Retention you control
Configurable retention for recordings and transcripts, deletion on request, and data processing agreements available for enterprise and partner accounts.
Platform security
Enterprise controls, standard issue.
Access control
Role-based permissions across marketing, support, finance, and admin functions.
Tenant isolation
Every business, partner, and client operates in a walled workspace — data, users, and wallets separated by design.
Audit logs
Sensitive actions are logged and reviewable, from template changes to key usage.
Hardened operations
Least-privilege access internally, environment separation, and continuous monitoring across the platform.
Regulatory compliance
Compliant is the only way we ship.
Communications in India is a regulated business — and we treat that as home turf, not an obstacle.
- TRAI and DLT for SMS and voice. Principal Entity guidance, header and template registration, promotional time-window enforcement, and DND scrubbing — handled by the platform.
- WhatsApp Business policy. Verified onboarding, approved templates, and policy-conformant messaging on the Business API.
- Calling consent and disclosure. Configurable disclosure prompts, consent logging, and calling-hour windows on outbound campaigns.
- Records that stand up. Full transcripts and dispositions on every AI call — evidence, not anecdotes, for audits and disputes.
AI safety, specifically
Generative agents need governance, so every agent runs inside an explicit guardrail policy:
- Scoped topics and forbidden claims per agent
- Mandatory disclosures where required
- Escalation to humans on defined triggers
- Hardening against attempts to talk the agent off-policy
- 100% of calls transcribed and auditable
Certifications and roadmap
Say it plainly, prove it steadily.
We publish our posture honestly: what's in place, and what's in progress — because enterprise trust is built on precision, not badges.
In place
- Encryption in transit and at rest
- DLT-compliant delivery for India
- DPDP-aligned, GDPR-ready practices
- DPAs for enterprise and partners
In progress
- ISO 27001 certification program
- Expanded penetration-testing cadence
On the roadmap
- SOC 2 Type II attestation
- Sector-specific attestations as enterprise deployments require
Send us the security questionnaire.
We answer them for a living now. Get the posture document, the DPA, and straight answers from the team that built the controls.