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

# Core Concepts

Inworld’s Agent Runtime turns complex AI pipelines into simple, composable graphs you can build, run, and deploy in minutes.

You can interact with the Agent Runtime through two main tools:

* The [**Node.js Agent Runtime SDK**](https://www.npmjs.com/package/@inworld/runtime), which provides APIs for building, executing, and integrating graphs directly in your application code.
* The [**Inworld CLI**](/node/cli/overview), a command-line toolkit to quickly initialize Runtime projects, serve graphs locally, and deploy them to hosted endpoints.

## How it works

At its core, the Agent Runtime executes [**graphs**](/node/core-concepts/graphs), directed networks of [**nodes**](/node/core-concepts/nodes) connected by [**edges**](/node/core-concepts/nodes) that form an executable pipeline. Think of a graph as encapsulating your application’s logic — whether that’s a speech-to-speech conversational pipeline, a routing flow, or a chat experience with memory. Once you create a graph, you can then execute it by passing in inputs to get outputs.

It's super easy to start building with a library of [**built-in nodes**](/node/core-concepts/nodes#built-in-nodes) for common tasks—such as text generation, speech synthesis, and audio processing. You can also create [**custom nodes**](/node/core-concepts/nodes#custom-nodes) that implement any custom logic you wish.

Below is a visualization of a graph for a simple speech-to-speech pipeline that can be used to power a realtime voice conversation.

<img src="https://mintcdn.com/inworldai/xNv-cF085gNQytGR/img/nodejs/runtime/sample-graph.png?fit=max&auto=format&n=xNv-cF085gNQytGR&q=85&s=8d704c7a3c951b142354b206e7879b11" alt="" width="844" height="588" data-path="img/nodejs/runtime/sample-graph.png" />

Once you’ve created a graph, you can:

* **Use it directly** in your Node.js backend via the SDK, or
* [**Deploy it**](/node/cli/deploy) to a hosted endpoint and call it remotely from any client.

## Getting started

The best way to start building with Agent Runtime is by exploring our templates and guides to understand the core concepts and available components.

* Looking to get started quickly? Try the [quickstart](/node/quickstart).
* Want to build something from scratch? Start with the pre-built [Voice Agent](/node/templates/voice-agent) template, or explore other [templates](/node/templates/overview)
* Integrating into an existing project? Build and [deploy](/node/cli/deploy) a graph, and integrate it into your project with a single endpoint.
