Telnyx

Edge compute. AI Agent platform. Global communications.

Running on the same infrastructure for ultra-low-latency, AI-powered experiences.

[Telnyx @telnyx]

Most Voice AI calls still assume one caller and one assistant.
That breaks the moment a real workflow needs another human in the loop.

Like when a customer needs a specialist.
A rep needs an SE.
Two people need to agree on a time before anything gets booked.

You could wire this together with call control, webhooks, conference logic, STT, speaker tracking, and post-call workflows.
But then the hard part becomes invisible:

- Who is speaking?
- Are they talking to the assistant, or to each other?
- Should the assistant answer, or stay quiet?
- Can it still run tools while the humans sort things out?

We just shipped native multi-participant voice calls for Telnyx AI Assistants.
The assistant can invite another participant into a live call, keep speaker identity attached to the conversation, and keep using its configured tools while the call is happening.

Using Skip Turn, when two humans are talking to each other, the assistant can stay quiet for that turn.

Then when someone addresses it again, it can come back in, summarize, book the meeting, update a CRM, or take the next configured action.

Voice AI is moving from answering calls to coordinating live work.
Multi-participant voice calls are now live for Telnyx AI Assistants.

Read more about it here: https://t.co/akTLNR3x2o

[Telnyx @telnyx]

@PlatoCosmos @lilyyang169 @smallest_AI Will miss you!

[Telnyx @telnyx]

Dispute callers wait 15+ min just to start the resolution process. Voice AI agents verify identity, pull transactions, and walk customers through resolution in one call. PCI-DSS compliant. Faster for them, fewer escalations for you.

See it live:

#VoiceAI #AIAgents

[Telnyx @telnyx]

"Sorry, can you repeat that?" If your Voice AI can't handle background noise, that's the only answer your callers are getting.

Real calls happen in real environments — call centers, warehouses, cars, streets. If your voice AI only works in a quiet room, it's not production-ready.

Telnyx Flux STT cuts through the chaos and stays locked on the speaker, even when the real world won't quiet down.

See how it handles noisy calls without breaking cadence → https://t.co/B9JVmqiVZx

#VoiceAI #SpeechRecognition #Telnyx #FluxSTT #AIAgents

[Telnyx @telnyx]

@meme_scavenge Yeap

[Telnyx @telnyx]

𝟭/ Conversation Workflows are live for Telnyx AI Assistants. Design a voice call as a graph, not a single prompt.

[Telnyx @telnyx]

𝟯/ Build it in Mission Control or POST conversation_flow to the Assistants API.

Docs: https://t.co/JQ2zeWRoZK

[Telnyx @telnyx]

𝟮/ Nodes are steps, each with their own instructions, tools, and optional model/voice override. Edges route between them: LLM conditions (natural language) or variable conditions (deterministic, on values like a verified ID or call duration).

[Telnyx @telnyx]

@natinati1905 This feature handles that quite well

[Telnyx @telnyx]

Most Voice AI calls still assume one caller and one assistant.
That breaks the moment a real workflow needs another human in the loop.

Like when a customer needs a specialist.
A rep needs an SE.
Two people need to agree on a time before anything gets booked.

You could wire this together with call control, webhooks, conference logic, STT, speaker tracking, and post-call workflows.
But then the hard part becomes invisible:

- Who is speaking?
- Are they talking to the assistant, or to each other?
- Should the assistant answer, or stay quiet?
- Can it still run tools while the humans sort things out?

We just shipped native multi-participant voice calls for Telnyx AI Assistants.
The assistant can invite another participant into a live call, keep speaker identity attached to the conversation, and keep using its configured tools while the call is happening.

Using Skip Turn, when two humans are talking to each other, the assistant can stay quiet for that turn.

Then when someone addresses it again, it can come back in, summarize, book the meeting, update a CRM, or take the next configured action.

Voice AI is moving from answering calls to coordinating live work.
Multi-participant voice calls are now live for Telnyx AI Assistants.

Read more about it here: https://t.co/akTLNR3x2o

[Telnyx @telnyx]

@shabbiritech19 @Teknium 👀

[Telnyx @telnyx]

@yuvinod We are looking to integrate with ASR providers that support Hinglish mroe naturally

[Telnyx @telnyx]

WhatsApp has 3.3B users, and now your business can answer calls inside it with an AI voice agent.

Tap a button → AI picks up → instant, 24/7, any language.

90-sec demo below.

Full setup guide with code here 👉

What would you build with this?#WhatsApp #VoiceAI #ConversationalAI #CustomerExperience (edited)

[Telnyx @telnyx]

Neon Health saved 2M+ human hours by replacing Twilio with Telnyx.

Their AI agents call insurers, navigate phone trees, and secure prior authorizations. On Twilio, latency compounded at every handoff. No single vendor owned the call path.

One platform now. Patients get treatments 2x faster.
https://t.co/9NKMjQNHLS
#VoiceAI #BreakFree #CustomerStory

[Telnyx @telnyx]

@soniox_ai Letsgo!

[Telnyx @telnyx]

Voice AI teams now have another STT option on Telnyx.
We just added @soniox_ai STT for real-time transcription workflows.

This matters because STT is one of those pieces you only notice when it gets things wrong.

If the caller switches languages, says a product name, gives an account number, or talks over background noise, the rest of the agent is only as good as the transcript it receives.

With Soniox now available on Telnyx, teams building voice agents get another model to test alongside the rest of their voice stack.

This is useful for multilingual agents, mixed-language calls, and workflows where names, codes, and domain terms matter.

You can read more about it here:

https://t.co/8sAHPjth6G

[Telnyx @telnyx]

Alibaba makes 5,000 outbound calls a day in Pakistan. Most in Urdu, and most go unanswered.
The problem: STT models confused Urdu for Hindi. TTS voices sounded foreign. Unattested numbers got flagged as spam.

Telnyx solved all three on one platform → https://t.co/fgpztRyPXC
#VoiceAI #CustomerStory

[Telnyx @telnyx]

@nickbruhman @slashapp Nice billboard back there 👀

[Telnyx @telnyx]

@rimelabs 's new Coda TTS is now self-hosted on Telnyx.

Developers can now use Rime’s newest multilingual TTS model through the Telnyx voice stack, across standalone Text-to-Speech API, Voice AI API, and the Voice AI platform.

Coda is available for real-time speech experiences via WebSocket streaming, REST API, and Mission Control Portal, with support for English, Spanish, French, German, Portuguese, Japanese, Arabic, and Hindi.

Read more about the new tts here:
https://t.co/BdHDsy0s2N

[Telnyx @telnyx]

@DeepgramAI @Coffee_and_NLP lets goo

[GMI Cloud @gmi_cloud]

Helloo SF, on May 26, we are hosting another Hard Problems Night for Agent Builders.

Hitting on agent memory, multi-agent, reliability, and infra with @lightfld , @telnyx , and @NVIDIAAI

🗓 Tue May 26 | 6–8:30 PM
📍 600 Townsend St, SF

Demos from @tinyhumansai , NodeMori, @Claude_Memory, Telnyx, and our team.

RSVP here: https://t.co/B1gzMV8Lht

[Telnyx @telnyx]

The PR was merged same day and it's already live in OpenClaw's production release.

If you're building voice AI agents, the infra choice just got simpler.

[Telnyx @telnyx]

OpenClaw just shipped native support for Telnyx Media Streaming in its voice-call plugin:
https://t.co/pn1MZ7TsT3

Developers building AI voice agents on OpenClaw can now run real-time, two-way conversational calls end-to-end on Telnyx infrastructure.

[Telnyx @telnyx]

Sub-200ms round-trip.

One platform, not a Frankenstack.