Book a demo Start building
No lag. No awkward pauses.

Voice agents that feel like talking to a human.

TalkShoc is the platform for building, testing and deploying AI voice agents on phone and web. Design calls on a visual builder, and connect a number in one click. Callers speak naturally, interrupt freely and get answers instantly, without the lag.

Start building free → Read the API
No credit card to start Bring your own keys Live in an afternoon

Works with the carriers and models you already use

  • Twilio
  • Telnyx
  • Vonage
  • Plivo
  • OpenAI
  • Gemini Live
  • Qwen-Omni
  • Deepgram
  • ElevenLabs
  • Any SIP trunk
Two engines, one platform

Pick the pipeline that fits the call.

Run a classic cascaded pipeline when you need full control over every component, or switch to native speech-to-speech when speed is the whole point. Same agent, same dashboard, one toggle.

Cascaded

full control

Mix and match the best transcription, language model and voice for your use case, and swap any of them without rebuilding the agent.

Caller→ STT→ LLM→ TTS
  • 20+ model and voice providers
  • Custom vocabularies and accent tuning
  • Text-level guardrails and redaction

Native speech-to-speech

human-like

Audio goes straight into a speech-native model and straight back out. No transcription hop, no vendor chain to configure, and conversations that feel human.

Caller→ S2S model→ Caller
  • Gemini Live, Qwen-Omni and more
  • Keeps tone, emotion and pacing
  • Zero pipeline setup
Everything in the box

From first prompt to production calls.

The parts you'd otherwise spend months stitching together: media transport, turn-taking, telephony, testing and analytics.

Visual agent builder

Map call flows as a node graph: greetings, branches, tool calls, transfers and fallbacks. No code needed, with code available when you want it.

Test in the browser

Talk to your agent from the dashboard before it ever touches a phone line. Iterate on prompts and voices in seconds.

Numbers in one click

Buy a number and we provision the SIP endpoint for you, or bring your own Twilio, Telnyx, Vonage or Plivo account. Inbound and outbound.

Natural turn-taking

Voice activity detection, barge-in and endpointing tuned for real phone lines, so callers can interrupt and the agent stops talking.

Post-call analytics

Recordings, transcripts, latency breakdowns, outcomes and cost per call, all searchable in one dashboard.

Multilingual by default

Handle many languages and mid-call accent shifts without separate agents for each market.

How it works

Live on a phone line in an afternoon.

No media servers to run, no WebRTC to debug, no carrier contracts to negotiate.

Design your agent

Write the prompt, pick a voice and an engine, and connect tools like your calendar, CRM or any webhook.

Connect a number

Buy a number in the dashboard or point your existing SIP trunk at TalkShoc. Web and mobile callers work too.

Go live and measure

Take real calls, watch latency and outcomes in real time, and improve the agent from what callers actually say.

For developers

A real API, not just a dashboard.

Everything you can click, you can script. Create agents, place calls and stream events from your own backend.

  • REST API with Python and Node SDKs for agents, numbers, calls and analytics.
  • Webhooks and tool calls so agents can read and write to your systems mid-conversation.
  • Bring your own keys for any model provider, or use ours and get one bill.
Get an API key → 
from talkshoc import TalkShoc

client = TalkShoc(api_key="ts_live_...")

agent = client.agents.create(
    name="Front desk",
    engine="speech-to-speech",   # or "cascaded"
    model="gemini-live",
    voice="aria",
    prompt="You book appointments for a dental clinic...",
)

call = client.calls.create(
    agent_id=agent.id,
    to="+14155550123",
)
print(call.status)  # "ringing"
Use cases

Built for the calls that run your business.

Agencies and product teams use TalkShoc to ship voice agents for their own customers, under their own brand.

Front desk & scheduling

Answer every call, book, reschedule and confirm appointments around the clock.

Lead qualification

Call inbound leads within seconds, ask the right questions, and hand hot ones to sales.

Customer support

Resolve routine questions, look up orders, and transfer to a human with full context.

Outbound campaigns

Reminders, renewals, surveys and follow-ups at scale, with every outcome logged.

FAQ

Questions, answered.

Can't find what you need? Email hello@talkshoc.net.

What's the difference between cascaded and speech-to-speech?

A cascaded agent transcribes the caller, sends the text to a language model, then speaks the reply with a text-to-speech voice. That gives you control over each step. A speech-to-speech agent uses one model that listens and speaks directly, which removes two hops so replies come back instantly, like talking to a person. You can switch an agent between the two at any time.

Can I bring my own API keys?

Yes. Add keys for your model, transcription and voice providers and those costs go straight to your own accounts. Or skip the setup and use ours with nothing to configure.

Can I keep my existing phone numbers?

Yes. Connect your Twilio, Telnyx, Vonage or Plivo account, or point any SIP trunk at TalkShoc. You can also buy new numbers directly from the dashboard.

Can I resell TalkShoc to my own clients?

Yes. TalkShoc offers a white-label dashboard and per-client usage tracking, so you can run voice agents for your customers under your own brand.

Which languages do you support?

Agents can speak and understand many languages, and can follow callers who switch accents or languages mid-call. Coverage depends on the model and voice you choose; the builder shows what each one supports.

How do I get started?

Request access and you can build and test an agent in the browser right away. No credit card needed to start.

Your first voice agent is a conversation away.

Build it in the browser today, put it on a phone line this week.

Start building free → Book a demo