Which channels can one agent run on?

A website widget carrying voice and text in one place, your phone line for inbound and outbound calls, and WhatsApp for conversations as well as alerts to staff. The agent is configured once and deployed across those channels, and it can be launched from any page on your site.

Can we pin a specific AI model per deployment?

Yes. Claude is the primary reasoning model, with GPT- and Gemini-class models in the routing layer and automatic failover between them, and a specific model can be pinned per deployment. That means no lock-in to a single model vendor.

Can the agent do things, not just answer questions?

Yes. After verifying identity it can look up account-specific information, create, update and search records mid-conversation, and book against real availability. Disclosure is tiered across public, account-specific and restricted information, reads are field-scoped, writes are permissioned, and every action is logged.

How is access controlled across departments?

Role-based access is scoped per department, agents and workspaces are separate, edits reach live through an approval workflow, and version history supports comparison and one-click restore. The audit trail is complete.

Where is our data stored, and is it secure?

Data residency is available in the US, EU and Middle East, with a single-tenant option and a zero-retention mode, and client data is not used for training. Data is encrypted with AES-256 at rest and TLS 1.2+ in transit. CETRAI is Cyber Essentials Plus certified and a SOC 2 Type II examination is underway.

What languages does CETRAI support?

75+ languages, detected automatically, including switching language mid-conversation. There are 30 distinct voices with regional accents, the agent can be interrupted mid-sentence, and a plain-language mode is available.

CETRAI

One AI Agent for Your Website, Your Phone Line, and WhatsApp

A complete AI agent platform: one voice and chat agent on your website, phone line and WhatsApp, with identity verification, governance and audit trail.

The complete AI agent platform — voice and chat, on your website, your phone line, and WhatsApp. Configure an agent once, deploy it across every channel, and keep identity verification, governance and a full audit trail in the same platform.

Who it's for

One platform, three ways to buy. Small teams start themselves in minutes; enterprise and government start with a demo and a pilot.

Small business — Start today, on your own. An AI receptionist that answers your website and your phone line, in 75+ languages, and books meetings against real availability. Website widget with voice and text in one place. Inbound phone answering and live booking. 75+ languages, detected automatically.
Commercial enterprise — Deploy across departments. Multiple agents and workspaces, role-based access scoped per department, and verified identity lookups against your own systems. Role-based access and edit-to-live approval workflow. Verified identity lookups with tiered disclosure. SSO via Entra ID, Okta, SAML and OIDC with MFA.
Government — Answer residents at any hour. Answers grounded in your published content with per-item source attribution, a complete audit trail, and data residency you choose. Complete audit trail and knowledge base versioning. US, EU and Middle East data residency; single-tenant option. Awarded a position on Montana's statewide AI contract through competitive procurement.

What the platform does

The same agent across every channel, acting inside your systems after verifying identity, with governance and an audit trail around every change.

AI voice and chat on your website, phone line and WhatsApp — Configure the agent once and deploy it to the channels your customers already use. The same knowledge, guardrails and action permissions apply wherever the conversation happens. Website widget carrying voice and text in one widget. Phone line, inbound and outbound. WhatsApp conversations, plus WhatsApp alerts to staff. Launched from any page on your site.
It acts, it does not just answer — The agent verifies identity before it discloses anything account-specific, then works inside your systems during the conversation. Disclosure is tiered — public information, account-specific information, and restricted information are separate levels — and the agent never confirms or denies details as a way around verification. Verified identity lookups. Tiered disclosure: public, account-specific, restricted. No confirmation-by-denial. Record creation, update and search mid-conversation. Live booking against real availability. Field-scoped reads, permissioned writes, and a full action log.
Governed the way regulated buyers ask for — Changes are reviewable, reversible and attributable. Edits move to live through an approval workflow, access is scoped by role and department, and the audit trail is complete. Version history with comparison and one-click restore. Edit-to-live approval workflow. Role-based access scoped per department. Complete audit trail. Multi-agent and multi-workspace.
The knowledge engine — Give the agent what it is allowed to know: paste text, upload files, or crawl your websites. Scanned documents are read automatically. Retrieval combines meaning and keyword matching, and every answer carries attribution to the item it came from. Paste text, upload files, crawl websites. Scanned documents read automatically. Meaning-plus-keyword retrieval. Per-item source attribution. Scheduled automatic refresh. Gap detection for questions the content cannot answer. Thumbs-down correction loop. Knowledge base versioning. Shared and isolated knowledge bases.
Compliance and reliability — The controls procurement asks about, summarised here and documented in full in the trust centre. Cyber Essentials Plus certified. SOC 2 Type II examination underway. HECVAT Light completed. WCAG 2.1 AA, with a VPAT available. AES-256 at rest, TLS 1.2+ in transit. US, EU and Middle East data residency. Single-tenant option. Zero-retention mode; no training on client data. 99.9% uptime SLA, geo-redundant. RPO 4 hours, RTO 8 hours. SSO via Entra ID, Okta, SAML and OIDC, with MFA.

See the complete platform

How to Set Up CETRAI AI Voice Agents

Get started with CETRAI AI voice agents in 4 simple steps

1. Register & Create Agent — Sign up and configure your AI agent in minutes. Choose voice, personality, language.
2. Train with Your Content — Upload your product catalog, FAQs, service descriptions, and website content.
3. Connect Your CRM — Link to Salesforce, HubSpot, Zoho, or Pipedrive. Leads sync automatically.
4. Deploy & Monitor — Copy one line of code and paste into your website. Watch conversations happen in real-time.

Video transcript

Welcome to CETRAI. Your visitors have questions. Right now, most of them wait — on hold, in a queue, or until Monday morning. CETRAI answers. Instantly, naturally, and in your visitor's own language — at any hour. Configure it once, and it works everywhere your people reach you — on your website, over the phone, and on WhatsApp. One connected AI front desk for your whole organization. And it does more than answer. It books appointments against real availability, updates your records mid-conversation, and hands off to the right person when a human is needed. Whether you're a small business, an enterprise team, or a public agency serving residents. Want to see how it works? Click "Try it live," then start the conversation — and speak with Kate yourself. Ask in any language, speak naturally, and hear how it sounds.

Frequently Asked Questions

The questions teams ask most before deploying a CETRAI voice or chat agent.

Which channels can one agent run on? A website widget carrying voice and text in one place, your phone line for inbound and outbound calls, and WhatsApp for conversations as well as alerts to staff. The agent is configured once and deployed across those channels, and it can be launched from any page on your site.
Can we pin a specific AI model per deployment? Yes. Claude is the primary reasoning model, with GPT- and Gemini-class models in the routing layer and automatic failover between them, and a specific model can be pinned per deployment. That means no lock-in to a single model vendor.
Can the agent do things, not just answer questions? Yes. After verifying identity it can look up account-specific information, create, update and search records mid-conversation, and book against real availability. Disclosure is tiered across public, account-specific and restricted information, reads are field-scoped, writes are permissioned, and every action is logged.
How is access controlled across departments? Role-based access is scoped per department, agents and workspaces are separate, edits reach live through an approval workflow, and version history supports comparison and one-click restore. The audit trail is complete.
Where is our data stored, and is it secure? Data residency is available in the US, EU and Middle East, with a single-tenant option and a zero-retention mode, and client data is not used for training. Data is encrypted with AES-256 at rest and TLS 1.2+ in transit. CETRAI is Cyber Essentials Plus certified and a SOC 2 Type II examination is underway.
What languages does CETRAI support? 75+ languages, detected automatically, including switching language mid-conversation. There are 30 distinct voices with regional accents, the agent can be interrupted mid-sentence, and a plain-language mode is available.

Popular guides

The most-read CETRAI research on choosing, pricing, and deploying AI voice agents.

Legal & trust

Review the terms, privacy, and security commitments behind the CETRAI platform.