Skip to main content
Agents are programs that accept input, use tools and LLMs, and produce output. They range from simple prompt-driven assistants to deterministic TypeScript workflows. There are two ways to get agents into your workspace:

​
Install from the Agent Hub

The Agent Hub contains pre-built agents ready to use. Browse available agents from your workspace, install one, and start a session. No code required.
  1. Open your workspace at app.guild.ai
  2. Go to Agents and click Add agent
  3. Browse available agents, select one, and install it
The browse view includes agents your organization has published internally (Team-installable) as well as agents listed publicly on the Agent Hub (On-Hub). Installed agents are available to everyone in the workspace and can be run manually or via triggers.

​
Publish to the Agent Hub

To share an agent with your organization (so others can install it), publish a validated version from the CLI. See Publish to the Agent Hub.

​
Build your own

Use the Agent SDK to build custom agents tailored to your team’s workflows. The SDK supports three agent types:

LLM agents

Pair a system prompt with tools. The LLM decides how to accomplish the task. The simplest way to build an agent.

Auto-managed state agents

Write deterministic TypeScript with automatic state management. Best for algorithmic, cost-predictable workflows.

Self-managed state agents

Event-driven state machines with parallel tool calls and explicit state control. Most flexible, but more complex.
The quickstart walks through building and publishing your first agent in five minutes.