SmartStraw: no-code voice agents for small businesses, powered by Vocal Bridge

Paste a Google Maps link, pick a tone, and SmartStraw deploys a live Vocal Bridge phone agent that answers questions, books appointments, and captures leads.

The problem: small businesses miss the calls they can't pick up

A grocery store, a salon, a dentist's office. The phone rings during the lunch rush, nobody can reach it, and the caller hangs up. That caller was a question to answer, an appointment to book, or a lead to capture, and it is gone. Hiring an answering service is expensive, and building a custom voice agent means standing up telephony, speech, and a knowledge base most owners have no time or staff to manage.

The gap is not that the technology does not exist. It is that the setup cost keeps it out of reach for the businesses that would benefit most. A storefront owner does not write code, provision phone numbers, or tune a system prompt.

What they built

SmartStraw collapses the whole setup of a phone agent into a single short flow. A business owner picks what the agent should do (answer questions, book appointments, capture leads), pastes a link to their business, and chooses a tone. SmartStraw does the rest, analyzing the business, drafting the agent's personality, training it on the scraped context, and provisioning a phone number, all in one pass.

The point is that a non-technical owner goes from a link to a working phone agent in about a minute, with no setup to learn. Voice fits because a storefront's customers already call, so the natural product is one that simply answers the phone for them.

How they use Vocal Bridge

SmartStraw is built on Vocal Bridge's Voice for your agent surface. SmartStraw is itself a product built on top of VB, a generator that programmatically creates and deploys VB voice agents on the owner's behalf. SmartStraw owns the funnel that a small business actually touches: URL scraping, prompt generation, personality presets, the agent dashboard, and the post-call KPIs and satisfaction metrics it derives from interactions.

Vocal Bridge owns the voice. SmartStraw did not build telephony, streaming speech-to-text, streaming text-to-speech, barge-in, or the real-time, sub-second turn-taking that made the live test sound natural. It hands VB a config and a knowledge base and gets back a deployed, callable agent.

import { createAgent } from "@vocalbridgeai/agents";

// SmartStraw generates this from a scraped Google Maps / website URL
const agent = await createAgent({
  name: "Raley's Assistant",
  voice: { personality: "professional" },
  instructions: buildSystemPrompt(scrapedBusiness), // address, hours, deli, etc.
  tools: ["answer_questions", "book_appointment", "capture_lead"],
  telephony: { provisionNumber: true },
});

console.log(agent.phoneNumber); // live, ready to take calls

Illustrative. The real SmartStraw pipeline generates richer prompts and wires Calendly for live booking.

What is genuinely hard here is keeping the agent responsive while it reasons over freshly scraped, unstructured business data. In the demo the agent said it was "pulling up the details" and kept the caller engaged with a follow-up question while it worked, the kind of conversational filler that keeps a real phone call from going silent. Vocal Bridge handles the low-latency voice loop so SmartStraw can focus on the business context.

"SmartStraw enables your businesses to create any kind of voice agent very easily." SmartStraw team (OnePrompt Agents)

What's hard about voice here

A non-technical owner can generate an agent in a minute only if none of the voice infrastructure lands on them.

• The agent has to hold a real phone conversation. A caller interrupts, asks a follow-up, and expects a natural reply, which means low latency and clean turn-taking. Vocal Bridge's real-time voice channel and barge-in handle that out of the box.
• A live phone number and speech pipeline is not a weekend project. Vocal Bridge provisions the number and runs the voice layer, so SmartStraw can spend its build on scraping the business and writing the prompt instead.

That is what lets the product promise a working phone agent from a single link.