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

# Introduction

> What EDN is, what makes it different, and how the Memory API, the Console and the Workspace fit together.

**EDN (Episodic Dialogue Network)** is a memory API for AI applications and
agents. It gives your AI long-term memory that keeps track of where every fact
came from, so your model only works from information it can trust.

Most AI models forget everything when a conversation ends. Simple memory layers
fix that by storing everything, including the model's own guesses, and then
feeding those guesses back to the model as if they were facts. EDN remembers
what your users say and what your documents contain, labels every memory with
its source, and decides what your model is allowed to see.

## What makes EDN different

> [card] Every memory has a source
>
> Each memory is labelled with where it came from: the user said it, the user
> confirmed it, it came from a document, or it was part of a conversation.

> [card] Your AI can't write facts into its own memory
>
> By design, anything your AI writes is stored separately and never given
> back to it as a fact until a user confirms it.

> [card] Corrections replace old facts
>
> When a user corrects something, the new fact replaces the old one. Nothing
> is silently overwritten, so the history stays traceable.

> [card] Answers can cite their sources
>
> When you work with documents in a project, EDN numbers each source so your
> model can cite it. If the documents don't cover the question, EDN tells
> your model to say so instead of guessing.

## How it works

**1. Add memory**

Your app sends conversations, documents and corrections to EDN.

**2. Build a SIP**

Before each model call, your app asks EDN for a SIP (Session Initialisation
Package): a short, labelled block of the most relevant memory.

**3. Call your model**

Your app adds the SIP to the prompt and calls the AI model of its choice.

**4. Save the interaction**

Your app sends the new exchange back to EDN, so new facts can be remembered
and the memory stays up to date.

See [How EDN works](https://docs.tensorpro.ai/concepts/how-edn-works.md) for the full picture.

## The parts of EDN

> [card] [Memory API](https://docs.tensorpro.ai/quickstart.md)
>
> The core product. A REST API that adds source-aware memory to your app or
> agent.

> [card] [Console](https://docs.tensorpro.ai/console/overview.md)
>
> Where you manage your account and API keys, and explore your memory
> visually.

> [card] [Workspace](https://docs.tensorpro.ai/console/workspaces.md)
>
> A chat app in the Console to try EDN with your own documents and model.

In the API, a separate scope of memory is called a **project**. In the Console,
each project appears as a **workspace**.

## Bring your own model

EDN works with any AI model: Anthropic, OpenAI, open-source or your own. EDN
never answers questions for you, and building the SIP uses no model. EDN uses a
language model only to classify some of what it stores, such as facts in a
user's message and corrections. See [How EDN works](https://docs.tensorpro.ai/concepts/how-edn-works.md).

## Access and base URL

EDN is in **private beta**, and access is by invitation.

```bash theme={"theme":"github-light"}
https://api.tensorpro.ai
```

This base URL is the private beta environment.

## Next steps

> [card] [Quickstart](https://docs.tensorpro.ai/quickstart.md)
>
> Save a memory and build your first SIP in five minutes.

> [card] [How EDN works](https://docs.tensorpro.ai/concepts/how-edn-works.md)
>
> The path from content in to context out.

> [card] [Explore the Console](https://docs.tensorpro.ai/console/overview.md)
>
> See EDN memory working before you write code.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.