Create an Agent with Studio
You don’t have to walk through the agent builder yourself. Tell Studio you want an agent and it interviews you, then creates it.
Start it
Open Studio and click the Create Agent card, or just type what you want:
“I need an agent that answers the phone for my dental clinic”
Studio takes it from there.
The five things it needs
Studio asks for these one or two at a time, conversationally — you don’t fill in a form.
| # | What it asks | Example |
|---|---|---|
| 1 | Agent name | ”Riley” |
| 2 | Company or business name | ”KD Dental Clinic” |
| 3 | The agent’s goal — what should it do on calls? | ”Book appointments and answer questions about our hours and services” |
| 4 | What information it should collect from callers | ”Name, phone number, preferred appointment time, whether they’re a new patient” |
| 5 | Inbound or outbound — does it answer calls or make them? | ”Inbound” |
Language, industry, and timezone are optional. Studio only asks about them if you bring them up, and otherwise uses sensible defaults.
The goal and the data-to-collect answers are what Dialora turns into the agent’s prompt. “Book appointments” produces a thinner agent than “Book appointments, and if we’re fully booked this week, offer the next two open slots and take a callback number.” Spend your effort here.
What you get
Studio creates the agent and shows you a result card with an Open button that takes you straight to it. Dialora generates the prompt and welcome message from your answers.
The agent is real and editable — everything Studio built is yours to fine-tune in the dashboard.
Changing it afterwards
You can keep going in the same conversation. Ask Studio to change something and it updates the agent for you:
“Make the welcome message shorter” “Switch it to a British voice” “Give it a 5 minute max call length”
Studio can change:
- Name, company, goal, data to collect, inbound/outbound type, language, industry, timezone
- The language model that answers on calls
- The voice (text-to-speech provider, model, and specific voice)
- Speech-to-text provider and model
- The welcome message — the first thing the agent says
- Max call length, speaking speed, and how long it waits for silence before replying
- Voicemail handling, ambient noise, filler words, and back-channeling
Changing the name, company, goal, data to collect, type, language, or industry regenerates the agent’s prompt and welcome message. If you’ve hand-tuned the prompt in the dashboard, that work is replaced. Make those changes first, then tune the prompt.
Voices and models are never guessed
If you ask for “something warmer” or “a British woman”, Studio looks up what’s actually available on your account and shows you the real options to pick from. It won’t invent a voice ID or set a model you don’t have access to — an unavailable value is rejected rather than silently applied.
Finishing the setup
Studio can take you the rest of the way in the same conversation:
- Attach a knowledge base — “attach my services KB to this agent” (set one up first)
- Link a phone number — “connect my main number to this agent”, or get a Buy a phone number button if you don’t have one yet
- Test it — “let me talk to it” gives you a button straight to the agent’s Playground
Then publish
Studio’s changes are saved to the agent’s draft like any other edit. Publish the version to put them in front of real callers — see Agent Versions, Draft and Publish.