Quickstart

Use this page when you want the shortest practical path to a working RocketAiFlow workflow.

The goal is to validate one real path, review the result, and confirm that the trunk, agent, call records, and monitoring are working before you expand the setup.

Complete the common setup first, then follow either the inbound path or the outbound path for the first test.

Before you start

Before you start, prepare only what is needed for the first test:

a Deepgram API key for the agent voice setup
SIP trunk credentials when you want to test real inbound or outbound calls
for outbound, one contact to add manually or a small CSV file to import after the campaign is created

Choose one path

Choose what you want to validate first. You do not need to complete both paths for the first quickstart run.

Inbound: Inbound AiAI Inbound Routing when an incoming call should reach the agent.

Outbound: AI Dialer FlowsDialer Campaigns when a campaign should call contacts.

Common setup

Configure the shared parts once: trunk, agent, prompt, optional functions, and a direct Phone test. After that, continue with AI Inbound Routing for inbound or Dialer Campaigns for outbound.

1. Configure and verify the trunk

Inbound routing and outbound campaigns depend on a configured trunk. For a quick setup, focus on the parts that make the trunk usable: Trunk Configuration, Registration, Auth, and Trunk Status.

Follow the screens in this order:

Trunk Configuration screen showing trunk enabled, trunk name, server URI, and shared user. — **Trunk Configuration.** Enable the trunk, then set the trunk name, server URI, and shared user supplied by your SIP provider.

Registration screen showing registration enabled, server URI, client URI, contact user, retry intervals, expiration, and line setting. — **Registration.** Enable registration when the trunk must receive inbound calls. Confirm server URI, client URI, contact user, retry intervals, expiration, and line settings.

Auth screen showing userpass authentication with username and password fields. — **Auth.** Keep auth type on `userpass`, then enter the username and password used to register with the provider.

Trunk Status screen showing registration Registered and endpoint Not in use. — **Trunk Status.** Check that Registration and Endpoint are OK.

2. Create the first agent

Create one agent with a clear goal, such as lead qualification, appointment intake, or routing to a human team.

To quickly test voices and languages, choose the language and voice provider, then write a short greeting message. The greeting is the first message the agent says during the call.

For voice, LLM, transcription, functions, templates, and advanced Agent settings, see Create Your First AI Voice Agent.

RocketAiFlow includes two preconfigured examples, Lead Qualification IT and Lead Qualification EN. Use them as starting examples, not production-ready agents. If you want to test a ready-made agent immediately, go to Phone.

Voice configuration screen showing name, Deepgram API key, language, voice provider, voice model, greeting, and contact template variables. — **Voice.** Set the agent name, language, voice provider, voice model, and greeting before the first test.

3. Configure the prompt

For the quickstart, the LLM prompt can stay minimal. Explain who the agent is, what it should do, and when it should stop or transfer the call.

If you use a preconfigured agent, the prompt is already included. For a detailed prompt structure, see Configure Agent Prompt.

LLM configuration screen showing provider, temperature, model, prompt instructions, and contact template variables. — **LLM.** Choose provider and model, then add a minimal prompt to start interacting with the agent.

4. Optional functions

Optional. Use functions only when the agent must perform an action: transfer a call, reschedule a contact, save lead data, or call an external API.

For this quickstart, you can skip them. To configure actions properly, see Add Functions.

Function setup screen showing ready-made functions and a custom API action option.

Test the agent with Phone

After configuring the agent, test it from the Phone section.

Phone is a WebRTC client that can register SIP users, place and receive calls, and run a direct AI voice agent test. For this quickstart, use Voice agent test mode to simulate an inbound-style call to the agent without depending on trunk routing.

open Phone
enter the password for the default SIP/WebRTC user 2000
select the agent you created in Voice agent test mode
start the call and check that the agent responds as expected

Phone section showing SIP configuration, voice agent test mode, keypad, status badges, volume, and auto-answer.

Inbound quickstart flow

Use this flow when the phone number on the inbound trunk should route incoming calls to the agent.

You can also set a concurrency limit. When the limit is reached, additional incoming calls are rejected and the caller hears the number as busy.

the inbound trunk is configured and registered
the agent exists with prompt and required functions

5A. Create AI Inbound Routing

Path to follow: Inbound AiAI Inbound RoutingAdd Inbound Then select only:

the agent you created
the trunk you created
whether the inbound route should be active
the maximum number of concurrent calls this agent can handle on the route

Inbound routing screen showing the phone number, selected agent, trunk, active switch, concurrency limit, and recording setting.

6A. Run an inbound test call

Call the phone number connected to the configured trunk and verify that the inbound route sends the call to the expected agent.

7A. Review the inbound result

After the test, open the Inbound Call Record and check the values saved for that call based on the options you selected. You can review the call transcript, recording when enabled, and timing data.

Inbound Call Record detail showing duration, timeline, recording, and call transcript.

Outbound quickstart flow

Use this path when the first workflow should place calls from a campaign against a controlled contact list.

Start this flow only after:

the outbound trunk path is configured and registered
the agent exists
the prompt and required functions are attached to the agent
a contact CSV is ready or one contact can be added manually to the campaign

5B. Create a first outbound campaign

Path to follow: AI Dialer FlowsDialer CampaignsAdd Campaigns Then configure a simple campaign for the first test:

select the agent you created
select the trunk you created
select the campaign timezone
set Concurrent Call Capacity to 1
leave Retry Minutes To Reschedule at the default value
save the campaign before importing contacts, because the import step needs a campaign to attach them to

Contacts are added in the next step because they need an existing campaign to attach to.

Outbound campaign setup for a small controlled test.

6B. Import contacts

After the campaign is saved, upload a CSV for the outbound campaign and set Usage to AI Dialer Flows. This is required when contacts must be called by the outbound dialer.

Map Phone as the required field. All other contact fields are optional, so keep the first import limited to the values the workflow actually needs:

contactId if you need to correlate the contact with an external CRM or database record
name and surname if the agent should personalize the conversation
data if you want to pass extra information to the agent during the call, such as city, plan, CRM notes, or lead type
priority if some contacts should be selected before others
scheduledAt if a contact should become callable only at a specific date and time
custom fields in data when the agent or API action needs business context

Before using the contacts in a campaign, confirm:

the expected number of contacts was processed
valid rows were inserted
failed rows are understood

7B. Run an outbound test call

Start the call and verify the result in Call Records.

After the first test with one contact, decide the campaign pace by selecting how many concurrent calls the campaign should place.

Outbound call records list showing campaign filters, answered and no-answer outcomes, timing, and details actions.

8B. Review outbound records and monitoring

After the test call, open the Call Record and check the outcome, transcript, and timing.

Call record detail showing recording playback and the call transcript.

9B. Monitor dashboard and performance

Use the Dashboard section to monitor campaign pace, active calls, call limits, dialing rate, and call outcomes in real time.

In Real-Time Agents & Trunks, check the trunk state: Not In Use means the trunk is available, while In Use means calls are running through it.

There is also a performance view for analyzing calls over a selected time range.

All Dialer dashboard panels are described in Dialer Dashboard Panels.

Real-time dialer dashboard showing campaign outcomes, dialing rate, calls volume, active calls, limits, and endpoint state.

What to check if something goes wrong

If the first test does not behave as expected, open the Troubleshooting page and follow the inbound or outbound checks.

the trunk shows Registered
the inbound route or campaign uses the correct agent
the result appears in Inbound Call Records or Call Records

TroubleshootingOpen the inbound and outbound checks when the first test does not behave as expected.

Next steps

Deployment OverviewReview the deploy path that connects trunk, agent, routing, campaign, and monitoring.AI Inbound RoutingGo deeper on inbound setup after the inbound trunk and agent are ready.AI Dialer FlowsGo deeper on outbound campaign setup after the outbound trunk and agent are ready.Configure Agent FunctionsReview call control, transfer, rescheduling, lead qualification, and custom API functions.Monitoring and VisibilityReview how the first workflow should be monitored once it is live.Processed ContactsReview contacts already handled by inbound or outbound workflows before opening the detailed call record.