You are here. See how this question connects to other ideas.
Select a node to open its page · Expand to read within the map
One question
How does an AI agent move from answers to actions?
Understand goals, tools, feedback and stopping conditions.
Here an LLM-based agent selects actions toward a goal, observes tool feedback and chooses what to do next. A predefined workflow can also call a model; one application may combine both patterns.
For a report task, an agent might find material, save a draft and check references. Sending it requires separate authority. Tool access is not unlimited permission.
Its place in an application
An agent can be an application feature or a shared service. Blocklet describes an application unit; agent describes a process of action.
Identity identifies the actor, data supplies working material, and an execution environment runs operations. These are useful infrastructure questions for persistent agents, not a universal three-part definition of an agent.
Sources and further reading
Check your understanding
Is an agent saying “done” enough to verify completion?
No. Inspect tool results and the target system’s actual state.
Continue along a learning path
- From AI to agents that actStep 2 of 5