The Easiest Way to Add AI Calling to Any App: Copy-Paste One Prompt

The short answer: the easiest way to add AI calling to an app is to let the AI coding tool you already use do the integration. Copy the Vistara AI integration prompt for your tool from vistaraai.in/integration, paste it into Claude Code, Cursor, ChatGPT, GitHub Copilot, Lovable, Bolt or any other AI coding tool, and answer its questions. The tool creates a voice agent, places a live test call to your own phone, and writes the code that starts calls from your product and receives every call's transcript, summary and extracted answers by webhook.

This guide explains how that works, what the prompt contains, what ends up in your codebase, and which tool to pick.

Three ways to add AI calling to a product

The third option works because the prompt gives the AI tool everything it would otherwise have to guess: exact endpoints, field rules, error codes, retry rules, webhook behaviour and a proven method for writing voice agents that sound human on the phone.

What is inside the prompt

• The complete Vistara AI API reference — creating and updating agents, starting calls, fetching results, rate limits and every error code with what your code should do about it.
• Webhooks and how to secure them. Vistara webhooks are not signed, so the prompt has your tool protect the endpoint with a secret token in the URL and confirm important results through the API before acting on them.
• The Voice Agent Builder — Vistara's method for writing agent prompts: spoken lines in native script (Devanagari for Hindi), numbers spoken as words, no repeated lines, one question per turn, and a spoken goodbye before every hang-up.
• A step-by-step workflow and an integration blueprint that the tool follows with you, one step at a time.
• Notes for the tool you picked — for example, Claude Code runs the API calls itself, while ChatGPT gives you commands to run.

Your API key is never part of the prompt. The tool reads it from the VISTARA_API_KEY environment variable and uses it only on the server.

What happens after you paste it

1. Account and key. The tool checks that you have an account at app.vistaraai.in and that your key works, by listing your agents.
2. The voice agent. Use an existing agent, improve one, or build a new one. The tool asks a few questions — your business, the goal of the call, language, voice, what to capture after each call — then writes the full agent prompt and creates the agent once you approve it.
3. A live test call. It repeats your number back, then places a test call, which is allowed at any time of day. Answer and talk like a real customer would. Afterwards it shows you the summary, sentiment, extracted answers, transcript and recording, and fixes whatever you didn't like.
4. The integration. It writes the code in your stack: a small client that starts calls, a webhook endpoint for results, a table that stores every call, and a job that catches missed webhooks.
5. A go-live checklist. A separate production agent, the webhook secret set, the key only on the server, an alert when the wallet runs low, calls scheduled inside the 09:00–20:00 IST calling window, a redial policy and an opt-out list.

What a call looks like in code

Starting a call is one request. This is the kind of call your tool will wire behind a button, an event or a scheduled job:

curl -X POST https://app.vistaraai.in/api/v1/calls/initiate \
  -H "x-api-key: $VISTARA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"phone":"+919876543210","agentId":"AGENT_UUID","name":"Ravi","metadata":{"lead_id":"L-1042"}}'

The response contains a callId. Shortly after the call ends, Vistara sends the result to your webhook:

{
  "event": "call.completed",
  "callId": "7b840e61-9c60-4966-9eb1-a1e12739fa82",
  "data": {
    "duration": 94,
    "summary": "Customer confirmed interest and booked a demo for Saturday morning.",
    "sentiment": "Positive",
    "answers": { "Was customer interested in demo?": "Yes", "Preferred demo day": "Saturday morning" },
    "meaningfulConversation": true,
    "recordingUrl": "https://…/call-7b840e61.mp3",
    "metadata": { "lead_id": "L-1042" }
  }
}

The answers object holds the data points you defined on the agent, keyed by each question's exact text, and metadata comes back unchanged so you can match the result to your own record.

Which AI tool should you use?

What it costs to try

The prompt is free. New accounts get ₹150 free credit with a GSTIN (₹50 without) — up to 200 calls, since unanswered calls are never charged. After that, connected minutes are billed from a prepaid wallet; see pricing.

When you don't need an integration at all

If calls don't have to start from your own product, you can skip the code: upload an Excel sheet and run a campaign from the dashboard (see how to use Vistara AI). And an AI tool still needs a developer's judgement — review the code it writes the way you would a colleague's, and run real test calls before customers hear the agent.