---
title: "Core concepts"
description: "The words Lookio uses, from workspace to credit, and the four query modes with what each answer costs."
canonical: "https://lookio.app/docs/get-started/concepts"
updated: "2026-10-24"
---

# Core concepts

## Workspace

A workspace holds everything: resources, assistants, tags, members, credits and API keys. People belong to one or more workspaces and switch between them from the sidebar. Credits and billing are per workspace.

There are two roles. An **admin** manages billing, members and API keys. A **member** works with knowledge, within the permissions an admin gave them. See [Members and permissions](/docs/platform/members-and-permissions).

## Resources

A resource is one piece of knowledge, stored once: a text, a file, a web page or a website.

| Type     | Value of `type` | What it is                                                      |
| -------- | --------------- | --------------------------------------------------------------- |
| Text     | `text`          | Text you paste, up to 100,000 characters.                       |
| File     | `file`          | An uploaded file, read by the right tool for its format.        |
| Web page | `url`           | One public address.                                             |
| Website  | `site`          | A site or a section of it. Its pages are read and kept in sync. |

A resource goes through `queued`, `processing`, then `ready` or `failed`. An assistant can only answer from a `ready` resource. See [Resources](/docs/platform/resources).

## Tags

A tag is a short word on a resource, such as `pricing` or `hr`. Tags organise resources, and an assistant can follow a tag to read every resource carrying it, now or later. See [Tags](/docs/platform/tags).

## Assistants

An assistant answers questions. It has:

- a **name**,
- **context**: what it is and who it talks to,
- **answer guidelines**: language, format and level of detail,
- **knowledge**: what it reads.

### Knowledge modes

An assistant reads from one of three sources at a time.

| Mode     | Value      | It reads                                                                |
| -------- | ---------- | ----------------------------------------------------------------------- |
| All      | `all`      | Every resource of the workspace, including those added later.           |
| Selected | `selected` | Only the resources you picked, one by one (`resource_ids`).             |
| Tags     | `tags`     | Every resource carrying a tag it follows (`follow_tags`), now or later. |

The lists of the modes you are not using are kept, so you can switch back without picking again. See [Assistants](/docs/platform/assistants).

## Conversations

In the app, a conversation is a thread with one assistant. Lookio stores it for you and only you see it. Over the API, every answer is kept in a conversation you continue with `conversation_id`: a person's own chats stay personal, and an API key's conversations, which can name the people who use your product with `end_user`, are readable by the workspace in its [logs](/docs/platform/logs). See [Conversations and end users](/docs/platform/conversations-and-end-users).

## Query modes

A query mode decides how an answer is produced and what it costs. Flash is the default.

| Mode      | Credits | How it works                                                                                                  |
| --------- | ------- | ------------------------------------------------------------------------------------------------------------- |
| **Eco**   | 10      | The cheapest. The answer may take longer to start.                                                            |
| **Flash** | 30      | Fast answers for chats and automations.                                                                       |
| **Deep**  | 80      | Splits the question, searches each part in parallel and answers with a stronger model. For complex questions. |

### What an answer costs

One price per answer, set by its mode (Eco 10, Flash 30, Deep 80), whatever the length of the answer. Every question that reaches the assistant is charged: an answer that finds nothing in your resources costs the same, and so does a greeting. There is no free answer.

The one thing that is refused for free is a question to an assistant that reads no resource. It is not sent to any model and holds no credit: the call fails with `assistant_has_no_resources` until you connect resources to the assistant.

Lookio reserves the price of the mode before the answer starts, then charges it once the answer is written. If your balance is below the price, the question is refused with `insufficient_credits`.

## Credits

Credits pay for answers. Importing, editing and reading are free. A workspace gets credits from its plan and from credit packs, and the balance is visible in the app, in `GET /v1/credits`, with `lookio credits` and in the `get_credit_balance` MCP tool. See [Credits and plans](/docs/platform/credits-and-plans).