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

# How Quivly Works

> Data flows in from your connected systems, gets unified into one customer model, and AI features work on top of that model.

Quivly has three layers. Data comes in, gets organized around customers, and AI acts on it.

## 1. Data in

You connect the systems you already use:

* **CRM** (Salesforce, HubSpot) — customers, contacts, opportunities, owners
* **Billing** (Stripe) — subscriptions, invoices, MRR
* **Support** (Pylon) — tickets, response times
* **Calls** (Fireflies, Fathom, Granola) — transcripts and summaries
* **Slack** — customer channel conversations
* **Product usage** — from your warehouse (Snowflake, BigQuery), PostHog, or pushed directly via the [Usage API](/integrations/product-usage/api-overview)

Quivly reads from these systems. It never writes back to your CRM, billing, or support tools on its own — outbound messages (like a Slack message or email from an agent) only happen through steps you configure and approve.

## 2. One customer model

Records from different systems are matched to a single customer — a Stripe customer, a HubSpot company, and a Slack channel all land on the same profile. You control how fields map through [field mappings](/field-mappings/introduction), and you can extend the model with [custom fields and objects](/data-models/custom-fields).

Think of it like a filing system that files every new document under the right customer automatically.

## 3. AI on top

Every AI feature reads from the same unified model:

| Feature                                      | What it does                                                                        |
| -------------------------------------------- | ----------------------------------------------------------------------------------- |
| [Health scores](/health-scores/introduction) | Scores each customer across revenue, usage, engagement, support, and market signals |
| [Ask Quivly](/product/ask-quivly)            | Answers questions across all connected data, in-app or in Slack                     |
| [Agents](/product/agents)                    | Runs automated workflows on triggers like record changes or health drops            |
| [Actions](/product/actions)                  | Drafts next steps for you to review and send                                        |
| [Notebooks](/product/notebooks)              | Writes customer documents (QBR prep, handoffs) from live data                       |
| [AI fields](/product/ai-fields)              | Computes field values from each customer's data on a schedule                       |
| [Skills](/product/skills)                    | Your reusable instructions that steer all of the above                              |

Developers can also point external AI tools (Claude, Cursor, ChatGPT) at the same data through the [Quivly MCP server](/developers/mcp-server).

## FAQ

<AccordionGroup>
  <Accordion title="Does Quivly modify data in my CRM or billing system?">
    No. Data integrations are read-only. Outbound activity (Slack messages, emails, calendar invites) only happens through agent steps or actions that you configure, and you can require human review before anything sends.
  </Accordion>

  <Accordion title="How do records from different systems get matched to one customer?">
    Quivly links records across systems using identifiers like domains and external IDs, and you can review and adjust the linking. See [cross-system linking](/field-mappings/cross-system-linking).
  </Accordion>

  <Accordion title="Do I need every integration for AI features to work?">
    No. Each feature works with whatever data you've connected — more sources just make answers and scores richer. Health score categories with no connected source are simply disabled.
  </Accordion>
</AccordionGroup>