Vibe as a Service: collaborative voice-first trip planning, powered by Vocal Bridge

Compass plans and books a whole trip by conversation, building a live itinerary on screen and pulling a second traveler into the call from a phone browser.

The problem with planning a trip together

Planning a trip is a group activity that most software treats as a solo one. Someone opens a dozen tabs, cross-checks flights against a hotel against a budget, then relays it over text to the people actually coming along. The people with opinions, the ones who want the Dallas Zoo on the list, are never in the tool when the decision gets made. They get a screenshot after the fact. Every added traveler means another round of copy-paste and another budget nobody is watching in one place. By the time the plan is agreed, prices have moved and booking is a separate chore.

What Vibe as a Service built

Vibe as a Service turns trip planning into a single spoken conversation that ends with a booked trip. You tell Compass where you are headed, who is coming, and what you want to spend, and it assembles the plan with you in real time. It searches flights and hotels in the background while you keep talking, brings a co-traveler into the call so their must-dos land on the same itinerary, then charges a wallet and books the pieces before the conversation ends.

A remote traveler joins from a phone browser through a link, with nothing to install, and Compass folds their input into the same shared plan. The itinerary is not a summary emailed later. It is the thing everyone builds together, live, as they speak.

How they use Vocal Bridge

Vibe as a Service runs on Vocal Bridge's Voice for your app surface, where the voice agent and the on-screen UI share one real-time data channel. When Compass adds a traveler, sets a budget, or locks a flight, the planning canvas updates and the budget bar moves in lockstep.

The split is clean. Vibe as a Service owns Compass, the itinerary state machine (Proposed, Sketched, On track, Locked), and integrations with Sabre, LandingAI, and PayPal. Vocal Bridge owns the real-time voice layer: WebRTC transport, the voice-to-voice model, transcript streaming, barge-in, multi-participant audio, and the sub-second responsiveness that makes a spoken plan feel like a conversation.

import { useAgentActions } from "@vocalbridgeai/react";

const { onAction, sendAction } = useAgentActions();

// agent to app: a locked booking updates the live itinerary and budget
onAction("lock_booking", ({ type, label, orderId, amount }) => {
  setItinerary((it) => it.map((n) =>
    n.type === type ? { ...n, status: "locked", label, orderId } : n
  ));
  setSpend((s) => s + amount);
});

// app to agent: a traveler edits a node on screen, tell Compass
const handleEdit = (id, changes) => sendAction("edit_node", { id, changes });

Illustrative. The production implementation carries richer trip and payment state.

What's hard about voice here

Dead air while a search runs. Fetching real Sabre prices takes seconds, and silence in a voice call feels broken. Compass keeps the line warm ("I'm already on it. Just waiting for the results to come through.") while the tool call resolves, using Vocal Bridge's real-time voice-to-voice stream and conversational fillers.

More than one person in the room. Once Sapna joins from her phone browser, two speakers interrupt and talk over each other. Multi-party WebRTC audio and barge-in are Vocal Bridge's job, so either traveler can cut in without derailing the agent.

Tying speech to live plan state. Budgets update, nodes go from sketched to locked, an order ID lands on the right line. Every spoken decision has to move the on-screen state and stay consistent, which is what the bidirectional data channel is for. The production lever is to make every transition explicit and recoverable, extending the demo's strong shared-plan pattern safely into real payments.

In the builder's words

"So now when your trip is done, you get everything planned. The payments also show up over here." Mukul Pathak, builder, Vibe as a Service