> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pranthora.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Integrations Overview

> Understand how integrations power agents to perform complex tasks.

# Integrations Overview

Integrations are the **heart of the system**, empowering agents to perform more complex and intelligent interactions.

We’ve designed integrations to be **easy to configure** — with just a few clicks, you can connect external systems and make them available for your agents to use. This creates a **unified integration layer** that lets your voice agents access external data and perform actions seamlessly.

***

## ⚙️ Unified Setup

From the **Integrations Library**, simply:

1. Choose your preferred integration type.
2. Click **Connect**.
3. Sign in to authorize access to the provider’s capabilities.

Once connected, the provider’s tools automatically become available for your agents to use — enabling actions like fetching data, sending updates, or interacting with third-party APIs.

This applies to **off-the-shelf integrations** (e.g., Google Calendar, HubSpot, or Notion) as well as **custom configurations** that you can build yourself.

***

## 🧩 Custom Integrations

If you’d like to set up a custom integration, Pranthora allows you to configure one easily. You can define:

* **Authentication method**
* **Endpoints and parameters**
* **Expected input/output schema**

This flexibility allows your agents to work with virtually any external system or API.

***

## ✂️ Output Schema Control

One of the key features of our integration system is **Output Schema Control**.
External APIs often return large and detailed payloads — but not all fields are relevant.

For example:

> When fetching events from Google Calendar, you might only need the **event name** and **start time**, not the entire event metadata.

With Output Schema Control, you can:

* Define a **custom schema** for what fields are returned.
* **Trim down** unnecessary data before it’s sent to the model or tools.
* Maintain efficient and focused interactions.

This ensures cleaner responses, faster execution, and easier downstream usage.

***