App Cloud basics
What is an AI agent?
An AI agent works toward a goal by using models, context, and available actions to choose and carry out steps. Its behavior depends on the boundaries you give it.
How it works.
A model can produce an answer from a request. An agent adds a process around that ability: consider the task, use relevant information, choose an allowed action, inspect its result and decide what comes next. Useful agents need a clear stopping point and a way to handle uncertainty.
A support agent might read a question, look at permitted account information and prepare a reply. Drafting a reply and sending it are different actions. Decide whether a person should review the message and what the agent should do if it cannot find a reliable answer.
- Give the agent a concrete goal.
- Limit context and actions to the task.
- Review consequential or uncertain results.
What this looks like.
- A research agent prepares a brief from supplied sources.
- A support agent organizes a question and drafts a reply.
- A workflow agent prepares the next step in a request process.
The pieces that matter.
Goal
Define a useful result and the conditions that mean the task is finished.
Context
Supply relevant documents, records and instructions that the agent is allowed to use.
Actions
Choose which tools and changes are available. Set boundaries for actions that affect other people.
Review
Inspect results and decide when the agent should ask for help or stop.
Before you begin.
Memory is not automatic understanding
Stored material must still be selected and interpreted appropriately.
Autonomy has a scope
More actions are not always more useful. Begin with a bounded task you can evaluate.
In your App Cloud.
Agents belong alongside apps in Your App Cloud. Use Exepad to shape a useful agent concept around models, data and services, and check the available actions and execution behavior for your intended workflow.
Explore the capabilities or find an app or agent to create. Plan allowances and usage are explained on Pricing.
A few useful distinctions.
Is a coding agent the same as an agent inside an app?
A coding agent helps create software. An app agent performs a task for the app’s users. The developer entry path and the running-agent experience have different purposes.
Should every agent run without review?
No. A draft or suggested action is often a useful outcome. Choose review points according to the task and its consequences.
Keep learning.
Your ideas, up and running
Your next idea has a home.
Build an app. Run an agent. Make it yours.
Start building