Glytos documentation
Glytos is a platform for building voice AI agents: an agent listens, thinks, speaks, and can take real actions in your systems while it does. You describe the agent once and run it on a phone number, in a browser, or as text, without rewriting it per channel.
Every conversation is stored as a full transcript with the prompt, model and voice it ran with, what each part cost, and where the time went. When an agent gets something wrong, that record is where you start.
Start here
Quickstart
Build a working agent and talk to it in a few minutes.
Core concepts
Agents, workflows, sessions, and the words the rest of these docs use.
Prompt agents
One prompt, one purpose. The fastest way to a useful agent.
Workflows
Branching conversations with tools, handoffs and guardrails.
Build with the API
The API covers building, running and observing agents, and there are official SDKs for Node, Python, PHP, .NET and Go, a browser SDK for web calls and chat, and a command-line tool.
Authentication
Create an API key and make your first request.
API reference
Every endpoint, generated from the OpenAPI spec.
SDKs
Node, Python, PHP, .NET, Go, the browser, and the CLI.
Webhooks
Get signed events pushed to your backend as calls happen.
If you are evaluating Glytos, the Quickstart is the shortest path to hearing an agent answer. It needs nothing installed.