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

# Settings

> What the live call does, and what it leaves behind: turn-taking, limits, webhooks, analysis, and retention.

The **Settings** tab is one panel in two halves: the rules that apply while a live conversation is
running, then what Vocily keeps and extracts once it ends.

<Note>
  **If you used Vocily before:** the **Outcome** tab has been folded into this one. Custom Analysis
  and Saved Artifacts are now the last two cards on this page. **Interruption Sensitivity** and
  **Response Eagerness** moved out, to the [Transcriber](/building-agents/transcriber#turn-taking)
  pill.

  **Transfer** and the closing message moved the other way - they are the **Transfer Call** and
  **End Call** tools on [Tools](/building-agents/tools).
</Note>

## Conversation Behavior

Shape how responsive and conversational the agent feels during a live call.

* **Noise Handling** - `Off`, `Standard`, `High Noise`, or `Strict Guard`, for progressively more
  aggressive background-noise handling. High Noise and Strict Guard expose a level slider from
  natural to aggressive.

Start with the default and change one control at a time. A noisy line may need stronger noise
handling.

<Note>
  **Turn-taking moved.** **Interruption sensitivity** and **Response speed** (previously called
  Response Eagerness) are now in the [Transcriber](/building-agents/transcriber#turn-taking) pill's
  gear, beside the speech settings - because "the agent cuts me off" is a speech problem and that
  is where you go looking. They are the same two values, shown in one place only.
</Note>

## Duration Limit

**Max Call Duration** is the safety ceiling for a live call. It accepts `1` to `30` minutes and
defaults to `10`. The runtime tries to wrap up shortly before the limit so the call does not end in
the middle of an answer.

**Call Limit Message** is what the caller hears when the limit is reached. The limit also protects
against runaway provider and model costs; it is not a replacement for a clear call flow.

## Silence Handling

**End on Silence** ends the call when neither side has spoken for the configured period. It accepts
`5` to `120` seconds and defaults to `25`.

**Silence Timeout Message** is the line played before the call ends.

Turn on **Enable silence reminder** to play a short check-in first. **Remind after** accepts `3`
seconds up to one second less than End on Silence, and defaults to `15`. Its **Reminder message** is
a single line - "Are you still there?".

Silence handling is useful for abandoned calls and dropped audio, but a very short timeout can cut
off callers who think before answering.

## Webhooks

Add HTTP endpoints that receive this agent's call events as they happen - signed, retried, and
logged. Use webhooks rather than an API tool to send data to your system *after* a call ends.

<Card title="Webhooks reference" icon="webhook" href="/developers/webhooks/overview">
  Event types, payload shapes, retries, and signature verification.
</Card>

## Custom Analysis

Custom Analysis turns each finished conversation into structured data you define - a summary, a
classification, a number, a list. The values are defined as **groups** in the workspace library;
this card chooses which groups this agent runs.

* The two counters show how many groups are attached and how many values that adds up to per call.
* Pick a group from the selector to **attach** it; detach it from the list below.
* **Manage groups** opens the workspace library, where the definitions live.
* The toggle stops all extraction for this agent without changing what is attached.

Detaching a group here never deletes it - the definition stays in the library for other agents.

Every new agent starts with **Default Analysis** already attached, which gives you a call summary,
caller sentiment, outcome, and language on every conversation without any setup.

<Card title="Custom Analysis: full reference" icon="clipboard-check" href="/building-agents/custom-analysis">
  Groups and analyses, editing Default Analysis, result formats, run conditions, regex extraction,
  testing, limits, and how results reach your webhooks.
</Card>

## Saved Artifacts

Choose what Vocily retains after each finished conversation. Both are on by default for new agents.

* **Save Call Audio** - keep a playable recording in [Conversation
  History](/conversations/history) for review, coaching, and call-quality investigation.
* **Transcript Storage** - keep the conversation transcript available for review, analysis, and
  follow-up.

Choose the settings that fit your team's review and privacy needs.

## Save and test

Select **Save Changes** after editing. Everything on this tab belongs to the **version you are
editing** - save the draft, then publish and make it live when you are ready.

Test at least these paths before going live:

1. Interrupt the agent while it is speaking.
2. Stay silent long enough to hear the reminder and the timeout behavior.
3. Let a call run to the duration limit and check the wrap-up.
4. Confirm the analysis values you expect appear on the conversation afterwards.

## What's next

* Turn transfer, hold, and closing behavior on in [Tools](/building-agents/tools).
* Choose which number or channel reaches the agent in [Deploy](/building-agents/deploy).
* Review what came out in [Conversation History](/conversations/history).