Voice agent control plane

Build and operate your agents

Start from a production workflow, test it against real conversations, and deploy without leaving this workspace.

Native S2S Agent Factory

One agent. Your choice of speaking model.

Build with Ultravox, OpenAI, Gemini, Grok, or Smallest AI Hydra right here. Keep your saved workflow, tools, and team when you change models. Configured Supafone keys are used by default; workspace provider keys take priority.

Read the S2S guide

Start here

Create a voice agent from a proven workflow

Choose a workflow, add your business details, then create and test the agent.

API access Loading your connected key…

Connect your Supafone key in Build agent to load saved agents and calls.

Agents0Connected agents

Phone numbers0Owned and assigned

Recent calls0Latest 100 records

Agents

0 agents

Connect a key or refresh to load agents.

Recent calls

0 calls

Recent calls will appear here.

Phone numbers & call files

Provision numbers or download recordings and transcripts when needed.

QA, usage, and billing

Review Supafone Supervisor guidance, call quality, account usage, and billing without leaving the workspace.

Outbound workspace

Build a campaign

Describe the outcome, attach an outbound agent, add consented recipients, then review the generated campaign YAML before anything is launched.

CampaignsDrafts and active outreach

Connect your Supafone key to load campaigns.

Campaign setupUnsaved draft

0 recipients

Start from a workflow

Campaign name

What should this campaign accomplish? Plain language is converted into the same portable campaign YAML used by the SDK and API.

Outbound agent Only the selected agent is assigned. Its prompt, tools, knowledge corpus, and voice remain unchanged.

Channels

Timezone

Max attempts

Start hour

End hour

Campaign YAMLEditable source of truth

Not validated

Validation is a dry run and makes no changes.

Launch starts real outreach. Supafone never launches from generation, validation, or draft save.

Agent workspace

Configure your voice agent

Draft

Live previewTest your agent

Test browser voiceTalk with the saved speaking model through its browser audio connection. Review available transcripts and Supervisor events; Hydra has no live transcript stream.

Checked at call start

Ready

Save your agent settings, allow microphone access, then start a browser call.

Browser tests use your saved agent and its selected voice connection.

Supafone SupervisorLive supervision events

READYStart the browser call to stream supervision events.

Set up your agent

Connect your account, choose the job and voice, then create the agent. You can test it below without leaving this page.

Your Supafone API key Connected key

Need an API key?

Your Supafone account works across hosted agents, WebRTC tests, and calling.

Speech to speech

Choose the speaking model

One shared workflow

Your model handles the conversation. Supafone keeps the call stages, tools, Manager, and specialist team.

Ultravox selected · managed default

Provider keys · Supafone managed or BYOK

Who supplies the speaking model key?

These settings apply to every agent using this provider on your account.

Ultravox supports the shared workflow and its existing external TTS catalog. Model changes apply to the next call after saving.

Provider credentials stay on Supafone’s server. Use configured Supafone keys by default, with your workspace key taking priority when connected.

Business name

Website domain

After creation, Supafone reads approved public pages, chunks and embeds them, and gives this agent a private grounded knowledge corpus. Your prompt and tool choices stay unchanged.

What should it do?

Inbound or outbound?

Voice

Selected voice

Browse included voicesConnect your Supafone key above to load the complete live catalog.

Advanced: use your own voice-provider account (BYOK)

Only use this when you need a private provider account or custom voice ID. The included Supafone catalog above does not need these keys.

Your BYOK voice-provider API key

This key is not used by Supafone managed voices.

Language pack

Keep one language, or let the same agent continue in up to four configured languages.

1 language

Greeting

Instructions

Create this agentUses the settings above and keeps you here for an immediate browser test.

Connect your key, configure the essentials, then create the agent.

Behavior

Agent name

Agent key

Where does it run?

System prompt

Workflow

The shared Agent Factory

Give every model the same plan.

Set the call stages and the evidence required to move forward. Manager can coordinate specialists; the server enforces the saved workflow.

Call stages

3–8 stages

Plan

Number of stages

Fields this agent can saveComma-separated field names. A stage can require these saved caller details before advancing.

A transition must follow the allowed next stage. Required fields must be saved, and required tools must return a successful result. Model advice cannot satisfy those checks.

Enable ManagerA separate reasoning agent proposes the next stage or assigns a configured specialist. Supafone validates each action against the current call state.

Tasks per call

Parallel tasks

Timeout (seconds)

Manager coordinatesChooses roles and proposes checked workflow actions.

Supervisor coachesOffers conversational guidance. Enable it under Features.

Enable specialist teamDefine focused roles and their allowed stages. With Manager enabled, specialists can also reason privately while the selected S2S model keeps speaking.

Specialist analysis is advisory. It cannot run arbitrary tools, invent completed actions, or bypass stage requirements.

Use the current date and timeProvide server time in the configured timezone when the agent reasons about appointments and follow-ups.

TimezoneUse an IANA timezone, or leave blank to use the account setting.

Runtime

Choose the speaking model in Setup. Supafone hosts the shared workflow across Ultravox, OpenAI, Gemini, Grok, and Hydra. The optional connections below are for an existing external agent platform.

Your agent platform API key

Phone carrier

Supafone-managed numbers require no carrier credentials. Add these settings only when you want to use your own carrier account.

Account SID

Auth token

Secret / token

From number

SIP trunk URI

SIP username

Voice

Speaking model & voice Native S2S models use their provider’s voices. Ultravox also supports the external TTS catalog and compatible language profiles below.

Advanced voice IDs

Managed provider voice ID

Custom BYOK voice ID

Features

Supafone SupervisorAdd live guidance, call review, and quality scoring. Turn it off when you only need the base agent runtime.

Connection

Record callsOpt in to saving call audio for playback and download.

Transcribe recordingsRequest post-call transcription of saved native audio.

Native post-call transcription requires recording and Supafone’s configured Deepgram key. Provider-supplied live transcripts are separate; Hydra provides no native live transcripts.

Knowledge base

Agent knowledge corpus

Website pages and uploaded documents are chunked, embedded, and retrieved for this agent. Sources remain isolated to the current Supafone account and agent.

Not indexed

0Sources

0Website pages

0Vector chunks

PendingRetrieval

WebsiteNot synced

Add a website domain under Basics.

DocumentsPDF, TXT, Markdown, or CSV · up to 5 MB

Create the agent, then upload policies, FAQs, price lists, or other approved material.

Vector libraryEmbedding and reranker status

Index a website or document to inspect its retrieval assets.

Test retrievalUses the same grounded retrieval path as the live agent

Corpus testCreate or load an agent, index a source, then ask a question.

Knowledge must be enabled under Actions before the live agent can query this corpus.

Actions

Attach the operational capabilities this agent may use. Every claimed action still requires a verified tool result.

Tools

IVR / DTMF navigationGive authorized outbound agents a bounded phone-tree navigation mode.

Knowledge uses the saved website and documents for this agent. MCP and custom tool credentials are managed through account integrations and are never written into generated code.

Use this agent in your app

Choose your language, then copy the generated configuration into your project.

acme-health-inbound-receptionistAgent key

Inbound phoneRoute

$0/moNumber add-on

Ready to use in your application.

Code matches the current agent settings.

Validate

Logs

Paste a Supafone key, then refresh or stream logs.

Logs are session-only in this browser.

Build and operate your agents

One agent. Your choice of speaking model.

Create a voice agent from a proven workflow

API keys

Team members

Agents

Recent calls

Phone numbers & call files

Phone numbers

Recent call files

QA, usage, and billing

Build a campaign

Configure your voice agent

Set up your agent

Choose the speaking model

Behavior

Workflow

Give every model the same plan.

Call stages

Runtime

Phone carrier

Voice

Your selected S2S voice

Features

Knowledge base

Agent knowledge corpus

Actions

Use this agent in your app

Validate

Logs

Call details

Choose a starting workflow