> ## Documentation Index
> Fetch the complete documentation index at: https://docs.a2v2.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Core Concepts

> The handful of ideas that explain how everything in A2V2.ai fits together.

A2V2.ai has a lot of features, but they're all built on a few core concepts.
Understand these and the rest of the product makes sense.

## Accounts & organizations

When you sign up you get an **organization** — a shared workspace that holds your
agents, knowledge, contacts, billing, and team. Everything you do happens inside
your organization, and data is isolated from every other organization on the
platform.

You sign in with **email** (a one-time passcode is sent to your inbox) or with
**Google**. You can invite teammates and assign them roles:

| Role      | Can do                                                          |
| --------- | --------------------------------------------------------------- |
| **Owner** | Everything, including billing and deleting the organization.    |
| **Admin** | Manage agents, knowledge, contacts, and team members.           |
| **User**  | Work with agents and content, without organization-level admin. |

## Agents

An **agent** is a single chatbot. Each agent has its own:

* **Knowledge base** — the content it can answer from
* **Instructions** — its persona, tone, and rules
* **Model & settings** — which AI model powers it and how creative it is
* **Appearance** — colors, avatar, and widget behavior
* **Conversations & contacts** — the people it has talked to

You can create as many agents as your plan allows, and duplicate an existing
agent to reuse its configuration as a starting point.

<Note>
  Agents are fully independent. Training one agent on a document does **not** make
  that document available to your other agents.
</Note>

## Knowledge base

The **knowledge base** is the set of **sources** an agent is trained on. A source
can be a file, a website, a YouTube video, or a Q\&A pair.

### How processing works

When you add a source, A2V2.ai doesn't just store it — it makes the content
*searchable* by the agent:

<Steps>
  <Step title="Ingest">
    Your file or URL is read and the raw text is extracted.
  </Step>

  <Step title="Chunk">
    The text is split into small, overlapping passages so the agent can retrieve
    just the relevant parts.
  </Step>

  <Step title="Embed & index">
    Each chunk is converted into a numerical representation (an *embedding*) and
    stored in a vector index. This is what lets the agent find the right passage
    for any question.
  </Step>
</Steps>

A source moves through statuses as this happens — from **unprocessed** to
**processing** to **Completed**. Only completed sources are used in answers. If
something goes wrong a source shows **Failed**, and you can
[reprocess it](/knowledge-base/manage-sources#reprocess-a-source).

### How answers are generated (RAG)

When a visitor asks a question, the agent doesn't guess from general knowledge.
It uses **Retrieval-Augmented Generation (RAG)**:

1. The question is used to **search your knowledge base** for the most relevant
   chunks.
2. Those chunks are handed to the AI model **as context**.
3. The model writes an answer grounded in that context — and A2V2.ai shows you the
   **sources** it used.

This is why answers cite their sources, and why adding better content directly
improves answer quality.

## Credits

Usage on A2V2.ai is metered in **credits**. You consume credits when:

* An agent **generates a message** (cost varies by model)
* A document is **processed** into the knowledge base
* **Voice** transcription is used

Your organization has two credit pools:

| Pool               | How you get it                                      | Resets?                     |
| ------------------ | --------------------------------------------------- | --------------------------- |
| **Plan credits**   | Included with your subscription each billing period | Yes — refilled every period |
| **Top-up credits** | Purchased manually or via auto-recharge             | No — they carry over        |

<Tip>
  Turn on **auto-recharge** in billing so your agents never stop answering when a
  busy period drains your plan credits.
</Tip>

If credits run out, agents stop generating new answers until the balance is
topped up or the plan period resets. You can watch your balance on the **Usage**
and **Analytics** pages.

## The build loop

Everything above comes together in the loop you'll repeat for every agent:

<Steps>
  <Step title="Create">[Create an agent](/building/create-agent)</Step>
  <Step title="Train">[Add knowledge](/knowledge-base/overview)</Step>
  <Step title="Shape">[Write instructions](/building/instructions) and [pick a model](/building/model-settings)</Step>
  <Step title="Test">[Sandbox](/testing/sandbox) and [Playground](/testing/playground)</Step>
  <Step title="Ship">[Customize](/customize/appearance) and [embed](/embed/install)</Step>
</Steps>