> ## Documentation Index
> Fetch the complete documentation index at: https://docs.eolasflow.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# VoiceFlows
> Create and manage AI calling configurations that link assistants to phone lines
VoiceFlows are the core of EolasFlow. Each VoiceFlow links an AI assistant to a phone line and defines how calls are handled. Think of a VoiceFlow as the complete configuration for a calling use case: who speaks, what line it uses, what the AI should achieve, and how to process the results.
## VoiceFlow Overview
Each VoiceFlow includes:
* An AI Assistant that handles the call
* A Phone Line used as caller ID (outbound) or receiving line (inbound)
* A Direction setting (Inbound or Outbound)
* Settings for greeting, instructions, tools, post-call automation, and analytics
## Creating a VoiceFlow
Click New VoiceFlow to launch the AI-powered creation wizard. You can describe what you need in plain language or pick a template.
Type a description of what you need (e.g. "Inbound reception for a dental clinic that books appointments") or select one of the industry templates: Inbound Reception, Outbound Sales, Appointment Booking, Medical Office, Legal Intake, Real Estate, Survey and Feedback. You can also click Start from Scratch for a blank VoiceFlow.
The AI asks about your preferred communication style. Pick from options like Warm and friendly, Professional and formal, Casual and upbeat, or Sympathetic and gentle.
Choose what this VoiceFlow should be able to do: Book Appointment, Answer Questions, Leave Message, Check Availability, and more. Select multiple goals.
The AI generates a complete VoiceFlow with an assistant, conversation flow, and configuration. Review the suggested assistant or select an existing one from your team. Click Edit to refine the AI-generated prompt before creating.
Click Create Workflow. Your VoiceFlow is created with the AI-generated flow, assistant, and settings. It starts in Draft mode. Toggle to Live when ready.
The AI wizard generates a complete conversation flow with nodes, transitions, prompts, and tool configurations. You can refine everything in the Studio and Settings after creation.
## VoiceFlow Settings
Once a VoiceFlow is created, open it to access the full settings across five tabs.
### General
Core configuration for this VoiceFlow:
| Setting | Description |
| ---------------------- | ------------------------------------------------------------------------- |
| Name | Internal label shown in dashboards and reports |
| AI Agent | The assistant assigned to this VoiceFlow |
| Direction | Inbound or Outbound (set at creation) |
| Internal Transfer Only | Toggle on if this VoiceFlow only receives transfers from other VoiceFlows |
| Phone Number | The phone line assigned to this VoiceFlow |
| WhatsApp Messaging | Enable AI-powered WhatsApp text conversations on this number |
| Objective | The primary goal of this VoiceFlow |
| Voicemail Detection | Detect and handle voicemails automatically on outbound calls |
| Ambient Audio | Play subtle background sounds during calls |
| Workflow Collection | Link an Activepieces collection to expose its MCP tools during calls |
The WhatsApp Messaging toggle only appears when a WhatsApp-capable phone number is assigned. See the [WhatsApp Messaging](/guides/whatsapp) guide for details.
### Instructions
The Instructions tab controls the global system prompt and safety guardrails.
The system prompt defines your assistant's personality and behavior for this VoiceFlow. It is prepended to every stage's prompt in the Studio, so use it for company-wide rules like brand voice, compliance requirements, and language preferences.
Use the Variables button to insert dynamic placeholders (contact name, company, etc.) that are filled in at call time.
Rail Guards are safety boundaries enforced across all stages. Define what the AI must never do. Each guard has a name and a block action. For example: "No Medical Advice", "Confidentiality", "No Competitor Mentions". Click Add to create new rail guards.
### Tools
The Tools tab manages global tools available to the AI in every stage. You can also add stage-specific tools on individual nodes in the [Studio](#voiceflow-studio).
Click Add Tool to see available tools:
| Tool | Description |
| ------------ | ---------------------------------------------------------------------- |
| Appointment | Check availability, book, list, or cancel appointments via the Planner |
| Send Message | Send an SMS or WhatsApp message to the caller mid-call |
| Web Search | Search the web in real time to answer questions during the call |
| Send Email | Send an email (confirmation, follow-up) during the call |
| Send DTMF | Send touch-tone signals to navigate phone menus (IVR systems) |
| MCP Tools | Tools from your linked Workflow Collection appear here automatically |
Click a tool to add it. Click the remove button to take it off. Tools added here are available in every stage of the flow.
If you have linked a Workflow Collection in the General tab, its MCP tools appear in the Add Tool dropdown automatically. This is how your Activepieces workflows become callable during live calls.
### Post-Call
Configure what happens automatically after each call ends.
| Feature | Description |
| ------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Auto-Create Tasks | Creates tasks from action items found during post-call analysis. Enabled by default. |
| Email Notifications | Sends an email with call summary, transcript, and action items. Add recipient emails and use Send Test Email to verify. |
| Webhooks | Sends call events to external systems (ERP, CRM, etc.). Select an endpoint, choose trigger events (Call Completed or Call Not Connected), add custom fields, and test. |
See the [Webhooks guide](/guides/webhooks) for payload format and the [Automation guide](/guides/automation) for building workflows.
### Analytics
Configure the post-call AI analysis that runs automatically after each conversation ends.
| Setting | Description |
| ------------------ | ------------------------------------------------------------------------------------------------------------------------- |
| Call Objectives | Questions the AI answers about each call, e.g. "Did the agent book a meeting?" |
| Outcome Categories | Custom categories for classifying calls, e.g. "Interested", "Not Ready", "DNC". Click Use Defaults for common categories. |
| Custom Questions | Specific questions to answer from the transcript, e.g. "What was the customer's main concern?" |
Start with one or two objectives. Once you see how the AI interprets calls, add more granular questions.
## VoiceFlow Studio
The Studio is the visual canvas you see when you open a VoiceFlow. Drag and drop nodes, connect them with transitions, and define what your assistant does at each stage of the call.
### How the Studio Works
Each node is a stage in the conversation. Edges between nodes are transitions triggered by outcomes. During a live call, the AI follows the flow in real time: it enters a stage, works toward objectives, and transitions when an outcome condition is met.
### Node Types
| Node Type | Description |
| ------------- | ---------------------------------------------------------------------------------------------------------- |
| Global | Settings applied to the entire flow. System prompt defaults and safety guardrails. |
| Greeting | The opening message when the call connects. First thing the AI says. |
| Conversation | A stage where the AI converses with specific instructions, objectives, and tools. The main building block. |
| End Call | Gracefully ends the call with an optional closing message. |
| Transfer Call | Transfers to a human agent, department, or external phone number. |
### Building a Flow
Click anywhere on the canvas to add a new node. Choose the node type.
Click any node to open its properties panel. Set instructions, objectives, and which tools are available in this stage.
Drag from one node's output handle to another's input handle to create a transition.
Click any transition edge to set an outcome label (e.g. "Interested") and an optional condition describing when it triggers.
Use width and height inputs in the properties panel. Dimensions persist across save and reload.
### AI-Generated Flows
Click Generate with AI in the toolbar. Describe what you want in plain language and the AI builds the complete flow with nodes, transitions, instructions, and outcome conditions.
### Flow Versioning
| State | Description |
| --------- | ----------------------------------------- |
| Draft | Being edited. Changes saved but not live. |
| Published | Live and used for calls. |
| Archived | Previous version, kept for reference. |
Click Publish to make your draft live. Only one version can be Published at a time.
### Testing Your Flow
Use the Test Call button in the Studio toolbar. Choose Web test (browser audio, no credits needed) or Phone test (calls your number). After the test, review the transcript and Flow Audit in Conversations.
### Flow Audit
After every call, the Studio shows a Flow Audit overlay on the canvas showing which stages were visited and which transitions were taken. Visited nodes are highlighted, skipped branches remain dim.
Review Flow Audits for your first 10 to 20 calls on any new flow. This quickly reveals whether transitions trigger at the right moments.
## VoiceFlow Types
Outbound sales VoiceFlows for prospecting, lead qualification, and scheduling demos. Pair with Outreach to reach large contact lists at scale.
Inbound support VoiceFlows answer customer calls 24/7. The AI resolves issues, answers FAQs using your Knowledge Base, and transfers to a human when needed.
Marketing VoiceFlows handle promotional outreach, event invitations, and re-engagement campaigns. Usually outbound.
Research VoiceFlows conduct structured surveys and feedback calls. Analytics objectives are especially useful here.
## VoiceFlow Status
| Status | Description |
| ------ | --------------------------------------------------- |
| Draft | Configured but not active. Calls are not connected. |
| Live | Active and handling calls. |
Toggle between Draft and Live at any time to pause without deleting your configuration.
## VoiceFlow Statistics
Open any VoiceFlow to see performance at a glance:
* Contacts Reached: total calls completed
* Response Rate: percentage of outbound calls answered
* Conversion Rate: percentage achieving the defined objective
* Average Call Duration: typical call length
## Inbound vs Outbound
Customers call your phone number. The VoiceFlow answers and the AI handles the conversation. Great for support, appointment booking, and after-hours coverage.
You call your contacts. Used with Outreach campaigns. Upload a contact list and the AI calls each one. For sales, follow-ups, reminders, and surveys.
A VoiceFlow can only be one direction. The direction is set at creation and cannot be changed. Create two separate VoiceFlows if you need both.
## Voice Widget
For inbound VoiceFlows, generate a widget key to embed a voice assistant on your website. Visitors click the widget to speak with your AI in real time, no phone call required.
See the [Voice Widget guide](/guides/widget) for the full setup instructions.
## Best Practices
Test your VoiceFlow using the Softphone before going live. Toggle AI Mode on, call the assigned number, and experience the call as a customer would.
After going live, review the first few calls in Conversations. Check transcripts, verify analytics categorization, and refine prompts.
Use the Analytics tab to understand patterns. Small prompt changes often make a large difference in call quality.
Before making automated outbound calls, ensure you have consent from your contacts. Follow TCPA (US), GDPR (EU), and local telemarketing regulations.
## Next Steps
Create and configure the voice persona behind your VoiceFlow
Run outbound calling campaigns using your VoiceFlow
Analyze call transcripts and Flow Audits after calls complete