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

# AI Employee Inbox

> Give each project an AI teammate with its own email address

The AI Employee Inbox is different from every other integration: there's nothing to connect and no account to authorize. Each project gets its own **AI employee with a real email address** — forward it anything about the project and it becomes knowledge the whole team can ask about.

<Note>
  Project inboxes are enabled per organization. If you don't see the inbox option, ask Brink to enable it for your org.
</Note>

## Set up a project inbox

<Steps>
  <Step title="Give this project an inbox">
    During onboarding, or any time from the sidebar account menu (**{Project} inbox**), set up the inbox: pick a name (Sam, Riley, Jordan, Avery, Quinn — or your own), an avatar, and a personality — **Balanced**, **Warm**, **Direct**, or **Concise**.
  </Step>

  <Step title="Share the address">
    Copy the **Inbox address** and share it with your team. Forward emails, CC it on threads, or have subs and vendors send documents directly to it.
  </Step>
</Steps>

<Warning>
  Anything sent or forwarded to this address becomes project knowledge — visible to everyone on this project. It is deliberately *not* a personal mailbox.
</Warning>

## Routines

The inbox page lists what your AI employee does on its own:

* **Auto-index inbound mail** *(always on)* — everything that arrives becomes searchable project knowledge, and attachments flow into the Document Center.
* **Auto-reply to teammates** *(on by default)* — answer people on this project directly from the inbox.
* **Morning briefing** *(off by default)* — a short daily summary of what came in overnight.

## The Activity feed

Every inbound message shows up in the **Activity** feed with the sender's tier — **Teammate**, **Known**, or **Unknown** — and what the employee decided: **Indexed**, **Created a task**, **Sent** (a reply), or **Ignored**.

## What the AI employee will and won't do

The AI employee is deliberately conservative:

* The **only** action it takes autonomously is an informational reply within an existing thread, to an address already on that thread.
* Anything consequential — sending new outbound mail, writing to a system of record, anything touching money or schedule — requires **human approval** and shows up as a proposed action on your task list.
* Machine-generated mail, reply loops, and mail between AI employees are filtered before any response is considered.
* The sender's tier caps what can ever be authorized — an unknown sender gets less than a teammate.