Getting Started
Build and ship your first AI agent in about five minutes.
What is CustomWeb AI?
CustomWeb AI is a platform for building AI agents that talk to your customers — over the phone, on your website, and over text. You design the agent's behavior, connect your tools and knowledge, and we run the real-time pipeline (speech-to-text → LLM → text-to-speech → telephony) for you.
Under the hood, every agent is an orchestration layer over three swappable modules — a transcriber (STT), a model (LLM), and a voice (TTS) — each of which can be any supported provider. You pick the providers; we run the streaming, failover, tool calls, and billing.
The four agent types
| Type | Channel | Best for |
|---|---|---|
| Call Agent | Phone (voice) | Inbound/outbound calls — reception, support, qualification |
| Workflow Call Agent | Phone (voice) | Complex, branching call flows built on a visual graph |
| Website Agent | Web chat + optional voice | An embeddable widget for your site |
| Text Agent | SMS | Two-way text conversations |
Quick start
- Create an agent — Dashboard → New Agent. You only need a name to start; everything else is configured in the builder.
- Set the prompt + model — open the agent, edit the System Prompt, and pick a model in the Models tab.
- Add knowledge or tools (optional) — upload documents in Knowledge, connect actions in Tools.
- Test it — use the built-in Test panel to chat or simulate a call before going live.
- Publish & deploy — publish the agent, then attach a phone number (voice) or copy the embed snippet (website).
Tip · New accounts start with 1 free token (1 token = $1 of usage), enough to explore the platform and make a few test calls before you add credits.
Platform-hosted vs BYOK
Every provider (LLM, STT, TTS, telephony) runs in one of two modes:
- Platform-hosted — use our provider keys and pay with tokens. Nothing to configure.
- BYOK (bring your own key) — connect your own provider keys in Integrations. You pay the provider directly, so that usage spends no tokens; instead a flat orchestration fee applies ($0.06/voice-min, $0.006/message) to cover running the pipeline.