> For the complete documentation index, see [llms.txt](https://docs.dapta.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.dapta.ai/dapta-mcp/skills/voice-agents.md).

# Voice Agents

What Claude can do with your Dapta voice agents once connected: create, configure, and calibrate agents, then review how their calls perform.

Once Dapta is connected to Claude, you can manage your voice agents in plain language. Describe what you want, and Claude handles the voice-agent work in your workspace.

{% hint style="info" %}
When Claude changes an existing agent or its prompt, it shows you a preview first and waits for your confirmation before applying it. Reading and reviewing never changes anything.
{% endhint %}

## Build and configure

Ask Claude to set up a new voice agent and it walks you through a few quick steps: pick a starting template, choose a language, and share your company details. Claude writes the agent's prompt for you and creates it in your workspace, ready for a test call.

{% hint style="info" %}
**Creating an agent from a website.** If you share your website, Claude reads it with the web tools of the client you are in (claude.ai, Claude Desktop and Claude Code have them). If your client has no web access, Claude asks you for the company details instead.
{% endhint %}

**Try saying:**

* "I want to create a voice agent for my dental clinic."
* "Set up a new outbound cold-call agent in English."
* "Build me a receptionist voice agent for my company."

Once an agent exists, you can fine-tune how it sounds and behaves. Ask Claude to review the current setup or adjust things like speaking speed, how natural the voice sounds, how easily the agent lets callers interrupt, its webhook, or the details it extracts after each call. Every change is previewed before it goes live.

**Try saying:**

* "My agent speaks too fast, can you slow it down?"
* "What voice settings does my agent use right now?"
* "Make my agent's voice sound more natural."

You can also improve how the agent actually talks. Claude can look at recent calls, diagnose what is working and what is not, then draft a better prompt (for example, stronger objection handling or clearer pronunciation) and, with your approval, apply it to the live agent.

**Try saying:**

* "Improve my voice agent's prompt and apply it."
* "My agent doesn't handle objections well, fix it."
* "Update my agent so it reads phone numbers more clearly."

**Behind the scenes:** `create_voice_agent` creates the agent. `list_agents`, `get_agent` and `get_agent_complete_analysis` read it. `preview_update_voice_agent` then `commit_update_voice_agent` apply any change.

***

## Calibrate and review performance

Beyond setup, Claude can help you understand what each agent captures and how it is doing on real calls. Ask what an agent records or extracts, and Claude explains the details it collects on each call and how often those fields actually get filled.

**Try saying:**

* "What does this agent capture from calls?"
* "Which details are actually getting filled in?"

**Behind the scenes:** `get_agent_post_call_schema`, `find_input_variables`, `discover_variables_from_calls`, `analyze_variable_usage`, `aggregate_variable_results`.

When you want to know how an agent is performing, ask for the numbers. Claude pulls your call totals, connection and success rates, average duration, credits used, sentiment, why calls ended, and how calls break down by outcome.

**Try saying:**

* "Analyze my agent and give me a full report."
* "How many calls did I get this week and what was the connection rate?"
* "Categorize my last 100 calls by outcome."

**Behind the scenes:** `get_call_analytics`, `get_agent_call_statistics`, `list_calls`, `get_call`, `search_calls_by_phone`, `list_call_examples_by_category`, `export_calls`.

You can dig into the human side too. Claude can summarize the thumbs-up and thumbs-down votes and written comments people leave on calls, show satisfaction over time, and surface the calls that got negative feedback.

**Try saying:**

* "What's my agent's satisfaction rate?"
* "Show me recent negative feedback and comments."

**Behind the scenes:** `get_feedback_summary`, `get_feedback_trends`, `get_comment_sentiment`, `list_negative_feedback`.

To hear what was actually said, ask Claude for your call recordings. It lays out the calls that have a recording in a clean table with the key details and a link you can open to listen back.

**Try saying:**

* "Show me the recordings from my last 20 calls."
* "Let me listen to the calls from February 3rd."

**Behind the scenes:** `list_calls_with_audio`.

Claude can also tell you the best times to reach people. It looks at your call history to find the strongest hours and weekdays to dial, and the windows to avoid.

**Try saying:**

* "What are the best hours to call for my sales agent?"
* "Which days of the week get the highest connection rates?"

**Behind the scenes:** `get_hourly_call_distribution`, `get_weekday_call_distribution`, `temporal_heatmap`, `temporal_hourly_by_org`.

Finally, Claude can find your sales-ready leads. It first works out what "qualified" means for that specific agent, then pulls the matching leads and shows you a simple funnel from total calls to connected to qualified.

**Try saying:**

* "Find me the qualified leads from my outbound sequence."
* "Show me a conversion funnel for my sales agent's calls."

**Behind the scenes:** `discover_agent_qualification_schema`, `find_qualified_leads`.

{% hint style="success" %}
Not sure where to start? Tell Claude what you are trying to do in your own words. It picks the right approach and, for anything that changes an existing agent, confirms with you before applying it.
{% endhint %}

***

## Related

{% content-ref url="/pages/qt7i0FJbX1dz6R8QXEq3" %}
[Voice Sequences](/dapta-mcp/skills/voice-sequences.md)
{% endcontent-ref %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.dapta.ai/dapta-mcp/skills/voice-agents.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.