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

# Isolation and multi-tenancy

> How Adopt keeps each client's data walled off from every other client's — the guarantee your own clients' confidentiality rests on.

Client confidentiality isn't a feature you switch on in Adopt — it's how the platform is built. This page explains the guarantee and why it holds.

## The guarantee

Every client (workstream) in Adopt is an **isolation boundary**. Data, files, tables, and results inside one workstream are inaccessible from any other. A person or agent scoped to Client A cannot reach Client B's data. This isn't a preference setting — it's structural.

## Why it's on the client, not the process

Isolation lives on the **workstream** (the client) and flows down to everything inside it — including each engagement (process). That's deliberate: a client's confidentiality shouldn't depend on how many processes you happen to run for them. Add a second process for a client and it inherits the same boundary automatically.

## What this means in practice

* A reviewer who works on one client's files literally cannot open another client's.
* An agent bound to a client reads only that client's Doc Stores and the Data Store rows it's been granted.
* Bulk-provisioning 300 clients creates 300 independent boundaries; there's no shared pool to leak across.

## For regulated work

This isolation, combined with the always-on [review gate](/core-concepts/human-in-the-loop) and recorded approvals, is what makes Adopt suitable for regulated finance and accounting work where client separation is a professional obligation.

## Next steps

1. [Workstreams and Engagements](/core-concepts/workstreams-and-engagements)
2. [Roles and permissions](/security-access/roles-and-permissions)
3. [Access scoping](/security-access/access-scoping)