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

# How to get help

> When something isn't working the way you expect, here's the fastest way to reach the Cura team — and what happens when a conversation needs a real person instead of the AI.

Most days, Cura runs quietly in the background. But when something looks off — a channel won't connect, a message didn't send, the AI replied in a way you didn't expect — you shouldn't have to guess. This article shows you how to reach us, what to include so we can help quickly, and how a conversation gets handed to a human when the AI can't resolve it on its own.

***

## Reach the Cura team

The quickest way to get a person is email.

* **Email:** [support@cura.so](mailto:support@cura.so)
* **The Support link** in the top navigation of these docs opens a new email to that same address.

<Tip>
  Write to us from the email address on your Cura account. It helps us find your **Workspace** faster and reply to the right person.
</Tip>

***

## What to include in your request

The more we know up front, the fewer round-trips it takes to fix things. When you write in, try to include:

<Steps>
  <Step title="Your workspace or branch">
    Tell us which **Workspace** (branch or location) you're working in. If you run more than one, the exact name saves us guessing.
  </Step>

  <Step title="The channel">
    Say which **Channel** it involves — LINE, Instagram, Messenger, and so on. Many issues are specific to one channel's rules.
  </Step>

  <Step title="What you saw">
    Describe what actually happened in plain words: the exact message or error text, roughly when it happened, and what you expected instead.

    <Tip>
      A screenshot is worth a paragraph. Include one of the screen or error whenever you can.
    </Tip>
  </Step>
</Steps>

<Note>
  Before you write in, it's worth a quick look at the [Frequently asked questions](/faqs/frequently-asked-questions) — the most common issues, like a channel showing as disconnected or the AI not replying, already have step-by-step fixes there.
</Note>

***

## When the AI hands a conversation to a human

Inside your **Inbox**, your AI is built to know its limits. When it can't resolve a customer's question on its own, it doesn't keep guessing — it **escalates**, handing the conversation to a person on your team. This is the graceful fallback for anything the AI can't finish itself.

Your AI escalates automatically when:

* **A customer asks for a person** — "Can I talk to someone?"
* **It can't help** — the question is outside what it knows or can do.
* **Something goes wrong** — an error stops it from completing the task reliably.

When that happens:

* The AI **stops replying**, so it never talks over the person taking the conversation.
* The **Conversation** is flagged **escalated** and its priority is raised to **High**, so it stands out in the Inbox.
* It's **assigned to a teammate** automatically — Cura works down a fallback chain (the channel owner, then the workspace's escalation contact or first manager, then the account's contact, then an admin) so it never lands unowned.
* That teammate is **notified** by email and a push notification, with a link straight to the conversation in the Inbox.

<Note>
  If you run more than one branch, the AI may ask the customer which location they mean before it hands off — so the conversation reaches the right team.
</Note>

<Tip>
  You can decide who receives these handoffs, and add your own triggers (like a refund request or an upset customer). See [Setting an escalation contact](/team/escalation-contacts) and [Escalation: when the AI hands off to a human](/ai-agent/escalation-handoff).
</Tip>

***

### Related articles

* [Frequently asked questions](/faqs/frequently-asked-questions)
* [Escalation: when the AI hands off to a human](/ai-agent/escalation-handoff)
* [Setting an escalation contact](/team/escalation-contacts)
* [Taking over from the AI (and handing back)](/inbox/taking-over-from-the-ai)