---
id: KB-IN-010
url: https://app.codecontract.io/help/integrations/connecting-the-platform-to-your-assistant
idioma: en
categoria: integraciones
audiencia: desarrollador
nivel: avanzado
actualizado: 2026-08-13
tambienEn: [es]
relacionados: [KB-IN-002, KB-MA-002]
citadoPor: [KB-IN-014]
---

# Connecting the platform to your own assistant

_Letting an AI assistant of yours query and act on your cases._

**Responde a:** connect claude or chatgpt to the platform · mcp assistant integration · query cases from my assistant · give an external ai access

If your team already works with an AI assistant, it can query your cases directly instead of someone copying and pasting what is inside. It is a connection that is authorised and scoped, not open access.

## The three access levels

| Level | What it allows | When to use it |
| --- | --- | --- |
| Read only | Query and answer questions about what is there | Nearly always. It solves 90% of cases |
| Read and write | Also create and modify | When the assistant has to leave work done |
| Full | All of the above without restriction | Rarely, and with a very good reason |

> [!IMPORTANT]
> Start with read-only and stay there if it serves you. An assistant that can write can make mistakes writing, and what it does is recorded in the name of the organisation that authorised the connection — yours.

## What it genuinely solves

**En corto**

- Asking about case status without opening the platform.
- Cross-referencing what is here with what you hold elsewhere.
- Preparing answers to customer questionnaires with real data.

> [!WARNING]
> The connection inherits permissions: the assistant sees no more than the account that authorised it. If that account is an administrator's, the assistant sees everything. Authorise it with a scoped account, not your most privileged one.

> [!NOTE]
> Every query the assistant makes consumes, exactly as if a person made it. An assistant querying in a loop spends in a loop: check the breakdown in the first week.

## Before connecting

Decide which assistant, with which account and at what level. Those three answers written down somewhere avoid the conversation a year from now about who authorised what.

**Can I revoke it?**

Yes, at any time.

**Is what it queries recorded?**

Yes, like any access to data.

**Does it see other organisations' documents?**

No. Nothing crosses the separation between organisations.

## Ejemplos

**A team authorises the connection with an administrator account to move fast.**

- Redoes it with a scoped, read-only account

→ The assistant still answers what they needed and can no longer see payroll.

**The team wants to query files from their own internal assistant.**

- Connects the assistant through the supported route
- Limits the scope to the files that account can see

→ Queries happen where people already work, and the assistant sees no more than the person would.

**The in-house assistant returns data from other files.**

- Reviews which permissions it is connected with

→ The assistant's scope is the account's, not the company's.

**It is connected and nobody knows what it queries.**

- Checks the assistant's call log

→ Usage is auditable like any other access.

**The assistant asserts things that are not in the documents.**

- Checks which files the answers come from

→ Claims can be cross-checked.

**An assistant answer is shared with a client.**

- Reviews the text and its source before sending

→ What goes out carries your judgement.
