Quickstart

From an empty machine to a working voice agent, in order.

From an empty terminal to your first live call, in about 20 minutes. Three pages, meant to be read in sequence.

Total time is roughly 20 minutes, most of it Docker pulling images. You need Docker and one AI provider API key. Every stack runs on your own infrastructure — no call audio, transcript, or credential ever leaves your machine unless you choose a cloud model provider.

The path

Step	Page	What you end with
1	Prerequisites	A machine that can run the stack, and the accounts you need.
2	Install and run	Ten containers up, API answering on :8000, dashboard on :3000.
3	Generated secrets and defaults	Knowing what was generated for you and what to change.

Once the stack is up, build your first agent and place your first call from the dashboard — no terminal needed:

• Create your first agent
• Test and call with your agent

Prefer the API? See Recipes.

Before you start

You do not need a telephony account to try VoicEra. A websocket agent runs entirely in the browser and needs only an STT, TTS, and LLM key. Add telephony when you want real phone numbers.

Run make application-up, not a bare docker compose up. It generates SECRET_KEY, INTERNAL_API_KEY, and PROVIDER_AUTH_ENCRYPTION_KEY into .env; without them the stack starts misconfigured.

Where next

Once a call works:

• Architecture — what you just started
• Running a campaign — outbound at volume
• Security hardening — before anyone else can reach it