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

# MCP – Basics

> How you connect Docusnap365 to your AI tool over MCP, whose permissions it works with, and which eight tools are available.

Docusnap365 exposes your environment through the **Model Context Protocol
(MCP)**. Any MCP-capable AI tool you already use — Claude, ChatGPT, or
another — can then work directly with your Docusnap365 data: query the
inventory, have relationships explained, create assets. The same questions the
built-in [AI Assistant](/docs/en/assistant/overview) answers, your own tool
answers too — inside the tool you are already working in.

## Connecting

You add Docusnap365 as a connector in your AI tool and sign in with your
Docusnap365 account — no key, no copying credentials. Sign-in works just like
the web interface.

The address is your Docusnap365 address with `/mcp` appended, for example
`https://docusnap365.com/mcp`.

<Note>
  Where you enter the address in your AI tool differs by provider. The
  provider's own instructions apply.
</Note>

## Permissions

The AI tool acts with your own role: it reads and changes exactly what you
yourself are allowed to in Docusnap365, no more. Every change is attributed
to you.

## Tools

Eight tools are available:

| Tool | Does |
| - | - |
| `discover`, `list_schema`, `describe` | find out which data exists and how it is structured |
| `query`, `get` | query the inventory — filter, sort, read individual assets |
| `create`, `update`, `delete` | create, change, and delete assets, as far as your role allows |

Which fields a type carries is in the [data model](/docs/en/api/data-model).

<Note>
  Before creating or changing a record, the tool presents the details to you
  for confirmation instead of carrying them out unasked.
</Note>

<Note>
  If a tool rejects a call — say, because a required field is missing or an id
  is invalid — your AI tool gets a message it can use to correct and retry the
  call. For an unexpected error, it only learns that the call failed — repeat
  the question, or check the matter in Docusnap365 itself.
</Note>

<Warning>
  Scan jobs cannot be started through MCP. Scanning remains an action in
  Docusnap365 itself.
</Warning>

## Quota

Every tool call counts as one request against your subscription's quota,
regardless of how much data the tool queries for it. MCP and the REST API
consume the same quota. Establishing the connection and retrieving the tool
list do not count.

A call that the tool rejects, for example because a required field is
missing, has already been counted. If your AI tool corrects the call and
retries it, both calls count.

<Warning>
  Once the quota is used up, Docusnap365 runs no further tool calls and no
  further REST API requests until the next period begins. Your AI tool receives
  a message stating when requests are possible again.
</Warning>

If the quota temporarily cannot be checked, Docusnap365 does not run the call
either. Repeat the question later.

## Limits

The AI tool knows only what is captured in Docusnap365. Your data leaves
Docusnap365 toward the tool you use — that tool's data protection terms
apply for this path.

<Warning>
  Responses may contain errors. Check figures you rely on in the corresponding
  list.
</Warning>

## Related

What the built-in Assistant can additionally do is at [AI Assistant –
Basics](/docs/en/assistant/overview).


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