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

# How Supported.tech works

> Follow a customer request from conversation to approved action, confirmation and human handoff.

A Supported.tech agent does more than generate a reply. It connects a customer conversation to the information and actions needed to move the work forward.

## The request-to-result flow

<Steps>
  <Step title="A customer asks">
    The request arrives by phone, chat, WhatsApp, SMS or social DM. The agent identifies the customer, intent and missing information.
  </Step>

  <Step title="The agent checks approved sources">
    It uses only the knowledge, integrations and tools you have made available—for example a CRM, calendar, order system or inventory service.
  </Step>

  <Step title="The agent applies your rules">
    Instructions define what it may answer, which actions require confirmation, when it should stop and when a person must take over.
  </Step>

  <Step title="The next action happens">
    The agent can book, update, look up, qualify, route, notify or trigger another approved workflow.
  </Step>

  <Step title="The outcome is confirmed">
    The customer receives a clear result. Your team can review the conversation, tool activity and outcome together.
  </Step>
</Steps>

## Voice conversations

For a live call, the platform converts speech to text, sends the conversation and relevant context to the selected language model, applies your tools and rules, then returns the response through the selected voice. Call records can include the recording, transcript, extracted fields and outcome.

## Messaging conversations

For chat, WhatsApp, SMS and supported social channels, the platform keeps the same agent instructions and approved tools while adapting the response to the channel. A handoff should preserve the request and relevant context so the customer does not need to begin again.

## What you control

* The agent's purpose, tone and initial message.
* The models and voices available under your plan.
* Knowledge sources, integrations and credentials.
* Business hours, disclosures and channel behavior.
* Tool permissions, confirmations and action limits.
* Human handoff and escalation rules.
* Retention and access settings available to your account.

<Tip>
  Begin with one complete job. A narrowly defined agent is easier to test, safer to launch and faster to improve.
</Tip>