> For the complete documentation index, see [llms.txt](https://docs.artific.nl/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.artific.nl/en/for-administrators/tools.md).

# Tools and integrations

Everything that lets an assistant or a Toolbox element act on the outside world.

This section is more technical than the rest of the documentation. That is unavoidable: connecting an API cannot be explained without talking about methods, headers and authentication. If you are not the person who will do this, hand these pages to whoever looks after your integrations.

## What is here

| Page                                                       | What it covers                                                                   |
| ---------------------------------------------------------- | -------------------------------------------------------------------------------- |
| [API tools](/en/for-administrators/tools/api-tools.md)     | Calling your own or a third party's API from an assistant.                       |
| [Secret keys](/en/for-administrators/tools/secret-keys.md) | Storing the credentials those calls need.                                        |
| [API access](/en/for-administrators/tools/api-access.md)   | Letting your own systems call the platform, using service accounts and API keys. |

Two related pages live elsewhere because they are configured per assistant:

* [Tools](/en/for-administrators/assistants/tools.md): switching tools on and off for an assistant.
* [Connectors](/en/for-administrators/assistants/connectors.md): connecting to MCP servers.

## Choosing the right mechanism

| You want to…                                              | Use                                                                             |
| --------------------------------------------------------- | ------------------------------------------------------------------------------- |
| Let an assistant look something up in one of your systems | An [API tool](/en/for-administrators/tools/api-tools.md)                        |
| Connect to a server that already speaks MCP               | A [connector](/en/for-administrators/assistants/connectors.md)                  |
| Let an assistant run one of your Toolbox elements         | An element tool, from the assistant's Tools tab                                 |
| Let an assistant search your documents                    | A [search tool](/en/for-administrators/knowledge-base/search-tools.md)          |
| Let your own software drive the platform                  | A [service account with an API key](/en/for-administrators/tools/api-access.md) |

Prefer a connector when the target already supports MCP: discovery, per-action control and per-user authorisation come for free. Build an API tool when it does not.

## Two rules worth following

**Describe tools for the model, not for people.** The assistant decides whether to use a tool by reading its description. A vague description produces a tool that fires at the wrong times or never fires at all. Say what it does, what it needs, and when it should be used.

**Grant the narrowest access that works.** A tool is a capability you are handing to a system that decides for itself when to use it. Read-only endpoints, scoped credentials and blocked actions are not paranoia: they are the design.