Platform

Every step designed. Every conversation on record.

Design each conversation as a flow, connect your own systems and numbers, and choose how your voice agents run. This page shows how it all fits together.

Get started with RenggoTalk to sales
How a call runs

One call, from ring to record.

Follow one phone call from the moment it rings to the record it leaves in your call log.

Step 1 of 6

The agent picks up on your line.

Calls arrive on a Twilio number, a SIP trunk from your carrier or a PBX extension. The agent answers with the greeting you wrote, or places the call itself for outbound work.

Step 2 of 6

Your flow decides what happens next.

Each turn moves through the flow you designed. Logic splits route the caller by intent, described in plain language, or by values the agent has already extracted.

Step 3 of 6

It calls your systems mid-conversation.

Import an OpenAPI spec and its operations become tools. The agent checks availability, books the slot or looks up the order, and works the response into the conversation.

Step 4 of 6

Answers come from your documents.

Knowledge bases can be scoped to a single step of the flow. A question about visiting hours searches your visitor policy, not everything you've uploaded.

Step 5 of 6

People take over when they should.

Transfer cold, or warm: the caller waits while your colleague hears a spoken briefing of the call so far, then the lines are bridged.

Step 6 of 6

Every call lands in the log.

Transcript, tool calls, extracted variables and the reason the call ended, plus the audio for agents with recording turned on.

Voice engines

Two ways to run a voice agent. Chosen per agent.

Renggo isn't tied to one model. Pick the architecture that fits the call, and change it later without rebuilding the flow.

Real-time

Full-duplex

One speech-to-speech model listens and speaks in the same stream, so callers can interrupt naturally and turns feel conversational.

Gemini Live is the current default.

  1. VADSilero · Energy
  2. Speech-to-textDeepgram
  3. Language modelGemini · OpenAI · Claude
  4. Text-to-speechCartesia · ElevenLabs

Pipeline

Composable

Voice activity detection, speech-to-text, a language model and text-to-speech run as separate stages. Each has its own provider, chosen for language, voice, cost or vendor requirements.

Both engines sit behind one interface in the Renggo orchestrator. Flows, tools and knowledge bases are defined once, independent of the engine, so new model architectures can be added without rebuilding your agents.

Flow builder

Design the conversation. Don't script it.

Build flows on a canvas: drag in nodes, connect them and set up each step in a side panel. Operations teams can own the flow, and engineers get full control where they need it.

Node types

  • ConversationSay something and listen for the answer, with full prompt control.
  • Logic SplitBranch on intent in plain language, or on extracted values.
  • Extract VariableCapture a name, a date or an ID from what the caller said.
  • FunctionCall any operation from an imported OpenAPI spec.
  • Call TransferHand the caller to a person, cold or warm with a spoken briefing.
  • Agent TransferPass the call to another agent mid-conversation.
  • Press DigitSend keypad tones to navigate an external phone menu.
  • GlobalRules that apply at every step of the call.
  • End CallClose the call cleanly with a final message.

Around the canvas

  • Attach knowledge bases to a whole agent or to a single node.
  • Every publish is a version you can roll back to in one click.
  • Talk to the agent in your browser before a number is connected.
  • Run a single node on its own to see exactly what it does.
  • Start from a template and adapt it to your systems.
  • Build and edit flows from an AI assistant through Renggo's MCP server.
Your systems

Connect your systems. Keep your keys.

Agents work with your own APIs and your own model provider accounts, and your engineers can build flows from an AI assistant.

Your API, as agent actions

Import an OpenAPI spec and its operations become tools the agent can call mid-conversation, like checking availability or creating a booking.

Build flows from an AI assistant

Renggo's MCP server lets an AI assistant create and edit flows for you.

Your model keys, or ours

Bring your own model provider keys and pay the provider directly, or use our managed key.

Telephony

Your numbers. Your carrier. Your bill.

Renggo connects to the telephony you already have, for inbound and outbound calls. Twilio bills you directly for call minutes, with no markup from us.

  • Twilio numbers

    Paste your Account SID and Auth Token. Renggo lists your numbers and configures the voice webhook for you.

  • SIP trunks

    Bring your own carrier: register a trunk, origination URI and IP allow-list, and route its traffic to your agents.

  • SIP domains

    Register desk phones, PBXs and softphones as SIP extensions that ring straight through to an agent.

  • Browser

    Talk to an agent from the dashboard with your laptop microphone. No number needed.

From account to first call

  1. Connect your Twilio account.

  2. Choose a number you already own, or buy one in Twilio.

  3. Assign the agent that should answer it.

  4. Call it. When you hang up, the full record is in your call log.

Chat and Telegram

Your website chat. Your Telegram bot.

The same agents answer in writing. Add the chat widget to your website or connect your own Telegram bot, and every conversation runs on the flows, tools and knowledge you already set up.

Website chat

  1. Choose the agent that should answer, then set the widget's name, colours and greeting.

  2. Copy one snippet into your website, or send it to your web developer.

  3. Test the chat from the dashboard before it goes live.

  4. Publish. Every conversation lands in your log, and your team can take over when needed.

Telegram

  1. Create a bot with @BotFather in Telegram and copy its token.

  2. Paste the token into Renggo and choose the agent that should answer.

  3. Renggo connects the bot to your agent for you.

  4. Message your bot to try it. Every conversation lands in your log, and your team can take over when needed.

Outbound

Place calls, not just take them.

The same agents call out for reminders, follow-ups and callbacks. Start a call from the dashboard or trigger it from your own system, with the details the conversation needs passed in as variables.

From the dashboardChoose the destination, the caller ID and the agent, add variables and place the call.
From your systemsTrigger the same call through the API, for example when an appointment is booked.
Context, not scriptsVariables reach the flow as tokens, so the agent opens with the right name, time and reason.
Recorded when you chooseTurn recording on for the call and play it back from the call record.
Operations

See every call. Keep every record.

When something goes wrong on a call, you can see exactly what the agent heard, decided and did.

Call logTranscript with speakers and timestamps, every tool call with its inputs and response, extracted variables, channel and caller number.
Why it endedCaller hung up, agent completed, transferred, timed out or failed. Recorded on every call.
RecordingsTurn recording on per agent and play calls back from the call record.
Usage and costMinutes, tokens and cost for each call, rolled up by day.
Telegram alertsA summary card in your team's Telegram chat when each call ends.
ExportDownload call records as CSV or JSON for your own archive.
RolesAdmins, members and read-only viewers, scoped to your organisation.
FAQ

Questions about the platform

Anything else? Ask us in your demo.

Is Renggo built only on Gemini?
No. Gemini Live is the default real-time engine today. Any agent can instead run on the pipeline engine, with separate speech-to-text, language-model and text-to-speech providers.
Do I need a new phone number?
No. Connect the Twilio numbers, SIP trunks or PBX extensions you already have. You can also talk to an agent in your browser before any number is connected.
What's the difference between a managed key and my own key?
With a managed key, model usage is included in the per-minute rate. With your own key, you pay the model provider directly and Renggo charges only for orchestration.

See the platform working on your own use case.

Tell us how your customers reach you and what they ask most, and we'll walk you through Renggo in your demo.

Get started with RenggoTalk to sales