Stop building agents on human-first clouds. Agents operate communication infrastructure directly, provisioning voice, messaging, and compute at machine speed. No tickets, no portals. Just real systems your agents can hear, speak, and act on.
Build agile. Launch faster. Scale globally. All on one platform built for real-time engagement.
This is the 6th launch from Telnyx. View more
ClawdTalk
Launched this week
It's time to talk to your Clawdbot.
ClawdTalk gives your agent a phone number so you can call, text or WhatsApp it from anywhere in the world.
ClawdTalk is voice-first, so you can have actual conversations instead of being confined to chat windows. It's secure by design, your agent can only call and text you, and it's easy to set up, just add your phone number and start talking.
Start free with 10 minutes of voice and 10 messages per day.
Your Clawdbot, reachable anywhere. Powered by Telnyx.
Free
Launch Team
Telnyx
@fiona_mcdonnell Hi Fiona. How does ClawdTalk handle interruptions, context switches, or overlapping speech during a call?
Telnyx
@kimberly_ross Great question! ClawdTalk is built on Telnyx Voice AI, which supports real-time interruptions. You can interrupt the agent mid-sentence and it'll stop and listen, no waiting for it to finish. For turn-taking, Telnyx uses smart end-of-turn detection so it knows the difference between you pausing to think vs. finishing your sentence. Context carries over throughout the call since it's the same agent with the same memory. It's pretty seamless in practice.
what is the latency for whole cycle?
it will be really cool if it is around 1200 ms.
Owning the telco layer end-to-end is the part that matters most for hitting that number. Most voice agent stacks compound latency across STT, LLM, TTS, and then the network hops between them. Telnyx skipping the middleman on the carrier side shaves off a chunk most wrappers can't touch.
@kulsoom_awan1 we support voices from Rime, Minimax, and Resemble on launch day. Just go to settings - voice, and pick one that best suits your bot.
@kulsoom_awan1 it automatically sets the voice during onboarding, but once your bot is set up, you can go to settings and adjust it to fit your needs.
Looks great, would be fun to have a voice on my buddy client.. Is this supports only english or are there any other language options?
@flayzeraynx we launched with English. Other language support coming soon.
@dcasem Awesome, I'll be following for the language supports
Migma AI
The ~1200ms latency because you own the telco layer is the real differentiator here - most voice AI feels robotic because of those extra network hops. Giving agents actual phone numbers opens up so many use cases beyond web chat. Curious about how the voice providers (Rime/Minimax/Resemble) compare in real-world conversations - any noticeable differences in handling interruptions or speaking pace?
Wow, Telnyx is seriously cool! Love the concept of agents directly managing comms. Curious how ClawdTalk handles potential abuse or spam calls with that direct number access?
@jaydev13 We handle via a whitelist. No one can reach your clawdbot unless you whitelist it. When you instantiate calls, it can temporarily add them to your whitelist. Even when you whitelist, it understands that it's an external context and you can instruct your clawdbot on how your want it to lock itself down.
Voice-first approach for AI agents is a natural evolution from terminal and chat interfaces. The whitelist-based security model seems well-designed for preventing unauthorized access. Curious about the session state management - if a voice call gets disconnected mid-conversation, does the agent preserve context for when the user calls back, or does it require a fresh start?