Virtual machine capacity
CPU, memory, storage, and network sizing are finalized during the pilot based on expected call volume.
Never miss a TTY call.
TTY Agent answers TTY/TDD calls on your own voice network, using local AI for routine questions and a browser queue when a human should take over.
It registers to your PBX like a regular SIP extension, decodes Baudot tones into text, replies over the same call, and keeps call audio, transcripts, model inference, and knowledge-base content inside your infrastructure.
Staffing a TTY line around the clock is hard.
TTY calls still matter for accessibility and public-service obligations, but many organizations receive them infrequently. Dedicated staffing is expensive, yet missing a call creates a real customer-service and compliance problem.
TTY Agent gives regulated teams a practical middle path: an on-prem appliance that handles routine public questions with a local model and escalates anything private, complex, or human-requested to your staff.
Built for customer-owned voice networks
Routes from your PBX like a regular phone extension, without a transfer workflow.
Decodes Baudot FSK into caller text and sends replies back as TTY-compatible tones.
Answers only from the knowledge base you own and edit. Routine questions stay inside the appliance.
Private, complex, or caller-requested escalation lands in a browser queue for your team.
Audio, transcripts, inference, and knowledge-base content stay on customer-controlled infrastructure.
Designed for ADA and Section 508 alignment with optional transcript persistence and release manifests.
How it works.
Your phone system routes the call
TTY Agent looks like a regular phone extension to your PBX. Send TTY calls to it the same way you would route any other line.
The AI answers routine questions
A local AI model answers using a knowledge base you own and edit: hours, locations, and common questions. Nothing goes to the cloud.
Your team handles the rest
When a caller wants a person, or asks about something private or complex, the call moves to a simple browser queue your agents pick up.
Audit-ready by design.
- Local inference with no cloud AI dependency.
- Customer-controlled knowledge base on your network.
- Optional transcript persistence, disabled in one setting for privacy-sensitive deployments.
- Security hardening applied at build: default-deny firewall, hardened SSH, no default credentials, and narrow privileges.
- Build manifest with SHA-256 published per release so customers can verify the appliance image they import.
- Aligned with ADA and Section 508. NIST 800-53 moderate-baseline and NIST 800-171 control mapping available on request.
Technical requirements.
TTY Agent is built for an on-prem pilot with one virtual appliance, a PBX route, and a trusted-LAN browser admin surface. The exact sizing and SIP settings are validated during deployment.
PBX reachability
The appliance needs trusted-LAN network access to the PBX and a SIP extension or trunk destination for TTY calls.
Test path and operators
Bring a TTY endpoint or capable softphone, initial public FAQs, and operator access for live escalation testing.
Frequently asked questions.
Does the call audio leave our network?
No. Audio, transcripts, the AI model, and your knowledge base all stay inside the appliance, on your infrastructure.
Will this work with our PBX?
The appliance speaks standard SIP. If your phone system is RFC-compliant, it can route calls to TTY Agent. Vendor-specific details are validated during the pilot.
Can transcript storage be turned off?
Yes. A single setting disables transcript storage entirely for privacy-sensitive deployments.
Does the AI make up answers?
Out of the box, the agent only answers from the knowledge base you provide. If a question is not covered, the call moves to a human.
Start with one pilot line.
Validate the appliance with one PBX route, one TTY test caller, and a short knowledge base. Expand after boot, registration, call handling, and human escalation are proven in the customer environment.
Talk to Tenacity