One control plane for private voice AI.

Wordcab Platform is one API, one CLI, and one control plane for voice runtimes, models, deployments, and evaluations — across infrastructure your team controls.

Voice AI breaks when real engineering work starts.

Sample audio is easy. Production is where things slow down — cleaning data, benchmarking models, fitting the stack to the target environment, clearing security review, supporting it after launch.

Wordcab removes that drag. Runtime, fine-tuning, deployment, and control on one product path.

The full-stack "Voice OS"

Wordcab Platform

The delivery model for private voice AI. One API and CLI for service selection, deployment, fine-tuning, and operations.

Packaged Solutions

Industry packaging for healthcare, contact centers, financial services, revenue, and inference providers.

A control plane your platform team can actually operate.

The runtime is half the product. The other half — identity, inventory, observability, evaluation — is what most open-source stacks miss. It's also what turns a working demo into a system the on-call rotation can live with.

SAML + SCIM
Okta · Azure AD · Keycloak

Identity & access

One access surface across every service. SCIM provisioning, audit logs via webhook or syslog, per-environment RBAC.

Prom + OTel
no proprietary telemetry

Observability, standard

Every service emits Prometheus metrics and OpenTelemetry traces. Grafana dashboards ship in the chart. Route anywhere.

99.9% SLA
Production tier

Runbook-driven ops

Preflight, support bundles, upgrade paths, and rollback procedures. 24/7 Gold Support in Production and Sovereign tiers.

A/B + eval
held-out test sets

Model experimentation

Run new models and prompts against held-out test sets before they touch production traffic. Test suites live in the platform.

Built to fit within existing workflows.

01 — Endpoints

OpenAI-compatible runtime endpoints

A familiar integration path for chat, transcription, embeddings, and speech.

Application teams keep their existing SDK integrations. The difference is where inference runs: inside the customer boundary, not through a hosted endpoint your security team cannot audit.

02 — Control plane

API-first control plane

Deployments, models, agents, usage, and evaluations — inspectable and automatable.

Everything the platform manages is available through a programmable API. No hidden state, no opaque admin panels. Platform teams get the surface they need to fold voice AI into existing automation.

03 — Operations

CLI-oriented operations model

Built for teams that want infrastructure-friendly workflows, not an opaque admin box.

Operators manage the system from a terminal, not a web console. Scriptable, composable, fits existing runbooks. The CLI is the primary interface for deployment, troubleshooting, and routine operations.

04 — Observability

Metrics and diagnostics

Monitoring, diagnostics, and upkeep are part of the product — not an afterthought.

Health checks, structured logs, and diagnostic exports ship with every service. Troubleshooting does not require guesswork — or a vendor ticket for basic visibility.

05 — Agents

Agentic AI ready

Multi-step voice and text workflows with tools, policies, and human-in-the-loop checkpoints — inside infrastructure you control.

Agents are not a separate bolt-on. The same endpoints, control plane, and observability extend to orchestrated flows. Teams move from simple transcription to supervised automation without replatforming.

Licensed software, with engineering attached.

Wordcab ships as a container license in three tiers. Gold Support and custom engineering are built into Production and Sovereign — not quoted as side consulting. Volume and deployment-model specifics are handled on the call.

Pilot

Single-node evaluation

Up to ~500 concurrent streams · business-hours support · 30-day deployment window.

  • Wordcab Voice runtime, containerized
  • OpenAI-compatible endpoints
  • Helm chart + Docker Compose
  • Business-hours email support
  • Evaluation license — convert to Production anytime
Start a pilot
Sovereign / Airgap

Regulated and dedicated

On-site install · dedicated engineer · bespoke SLA · white-label available.

  • Everything in Production
  • Airgap, sovereign, or dedicated deployment
  • On-site install + validation
  • Dedicated engineer per deployment
  • Bespoke SLA with named escalation
  • White-label and OEM terms available
Contact sales
TCO anchor, in real numbers

A contact center running 200,000 calls/month at 10 min/call with stereo audio pays roughly $27,450/month to AWS Transcribe — or $35,250/month with Call Analytics. That's ~$330–$423k/year before redaction, language ID, or egress. A Wordcab Production deployment at that volume typically reaches payback inside 8 months and compounds from there, with zero audio leaving your boundary.

Frequently asked questions

How does Wordcab price?
Three container-licensed tiers — Pilot, Production, Sovereign. Production and Sovereign include Gold Support and a quarterly custom-engineering quota, not billed separately. Volume, deployment model, and SLA shape the number. See the pricing section above.
Is Wordcab a product or a services engagement?
A licensed product. Gold Support and custom engineering are packaged into Production and Sovereign so buyers get a system they can run — not a contractor that owns the runtime.
Why talk about a control plane at all?
Because private voice AI gets messy fast when deployments, models, transcripts, evaluations, and access controls all live in different places. The control plane is what makes the system operable — not just runnable.
Can audio and transcripts stay in our environment?
Yes. In customer-controlled deployments, sensitive data stays inside approved infrastructure boundaries.
Do we have to use one fixed model stack?
No. Model choice follows the deployment boundary, workload, and hardware profile.
Can we start with Wordcab Voice and add Think or Adapt later?
Yes. Start with the runtime. Add Think when reasoning or structured outputs become the blocker. Add Adapt when production fit becomes the main issue.

Own the runtime and the control plane.

If your team is choosing between building it, paying hosted APIs, or moving to private deployment — Wordcab gives you the control surface, runtime, and deployment model to run voice AI like real infrastructure.

Talk to an Engineer

We usually respond within one business day.

What are you building?

Or email us directly.