CETRAI

The Complete AI Agent Platform: One Login, Every Channel

Agent, knowledge engine, console and integrations in one platform: multi-model routing, verified identity actions, governance and a full 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.

What the platform actually includes

The agent, the knowledge engine, the console, the governance layer, omnichannel delivery and the integration layer are one platform behind one login — not separate products stitched together. The agent reasons on a multi-model routing layer with automatic failover, answers by voice or text on your website, your phone line and WhatsApp, verifies who it is speaking to before disclosing anything account-specific, and can create, update, search and book against your systems mid-conversation. Every change is versioned, approved and logged.

Complete, not assembled

One platform, one login. The agent, the knowledge engine, the management console, the governance layer, omnichannel delivery and the integration layer are parts of the same product, so there is no seam between the thing that answers and the things that control it.

  • Voice and chat agent
  • Knowledge engine
  • Management console
  • Governance layer
  • Omnichannel delivery
  • Integration layer

Multi-model intelligence, no vendor lock-in

Claude is the primary reasoning model, with GPT- and Gemini-class models in the routing layer. Failover between models is automatic, and a model can be pinned per deployment when a specific one has to be used.

  • Claude as the primary reasoning model
  • GPT- and Gemini-class models in the routing layer
  • Automatic failover between models
  • Per-deployment model pinning

One agent, every channel

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

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

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

Ask your deployment a question from Claude or ChatGPT

Staff query their own deployment in the assistant they already use — every conversation that ended badly last week, a specific transcript, what is trending in a department — with no report request and no export. The connection is made over MCP, is workspace-locked, admin-controlled and revocable.

  • Direct queries from Claude and ChatGPT over MCP
  • Workspace-locked, admin-controlled, revocable access
  • REST API
  • Signed webhooks
  • Custom tools pointed at any endpoint

Conversational intelligence

The agent meets people in their own language and register, and behaves sensibly when a conversation is difficult.

  • 75+ languages detected automatically, with mid-conversation switching
  • 30 distinct voices with regional accents, interruptible mid-sentence
  • Plain-language mode
  • Per-language disclaimers
  • Sentiment monitoring
  • Crisis awareness configured to your protective procedures
  • Guided screen assistance, with consent

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

Deployment

Implementation is white-glove and pilot-first, with a named project manager and written acceptance criteria, so the first deployment is scoped and signed off rather than guessed at.

  • White-glove implementation
  • Pilot-first rollout
  • Named project management with written acceptance criteria
  • Staff training and runbooks
  • Enterprise options: white-label, custom widget, enhanced compliance mode, SSO, custom voices, private dataset ingestion
  • CETRAI is available to Montana state agencies through the State of Montana Master AI Products and Services Contract.

Frequently Asked Questions

The questions buyers ask most about the CETRAI platform.

  • Which channels can one agent run on? A website widget carrying voice and text together, 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.
  • Which AI models does CETRAI use, and can we pin one? Claude is the primary reasoning model, with GPT- and Gemini-class models in the routing layer and automatic failover between them. A specific model can be pinned per deployment, so you are not locked to a single vendor's availability or roadmap.
  • How does the agent handle account-specific information? It verifies identity first, then discloses at the level that identity allows: public information, account-specific information, and restricted information are separate tiers. It does not confirm or deny details as a way around verification. Reads are field-scoped, writes are permissioned, and every action is logged.
  • What governance controls does the platform include? Version history with comparison and one-click restore, an edit-to-live approval workflow, role-based access scoped per department, a complete audit trail, and multi-agent, multi-workspace separation.
  • Where is our data stored, and is it used for training? Data residency is available in the US, EU and Middle East, with a single-tenant option and a zero-retention mode. Client data is not used for training. Data is encrypted with AES-256 at rest and TLS 1.2+ in transit, and the platform carries a 99.9% uptime SLA on geo-redundant infrastructure.

See the platform on your own content

Small teams can start on the website widget themselves. Enterprise and government deployments begin with a demo, then a pilot with written acceptance criteria.