How an agent is built
One pass, in order: where the agent’s knowledge comes from, then the four parts of the agentic core that decide what it does with it.
One agent, any channel, complete workflows.
The surface
The configuration surface itself: one agent, and everything the build path above sets, on a single screen.
The builder sets the boundary. The runtime holds it.
The left lane defines what the agent should know and do. The right lane shows the check or action that definition becomes while a conversation is moving.
One agent definition governs voice, SMS and email because configuration and execution belong to the same stack.
Guardrails, enforced at runtime
What you set as guardrails is what the runtime checks. Every sentence the model produces is checked before the caller hears it.
Onboarding
The team learns how you talk to your customers, then a Customer Success Manager tracks the numbers that matter to you.
“…it’s super helpful to have another set of eyes supporting you.”
Configuration becomes runtime law.
Approved knowledge, actions and handoff conditions become the boundaries checked during a live turn. The engine enforces them while it forms and speaks the answer.
Knowledge, intent, tools and guardrails enter one configuration. The runtime applies that definition across voice, SMS and email.
What the builder sets is what the runtime is allowed to say and do.
Try the live demo
Drop in your website. Alebex reads it, learns what you do, and calls you in under 30 seconds — fully trained, zero setup.