What are Agentic models?

1 · In one line

An agent plans, acts, observes the result, adjusts its approach, and repeats until the task is complete.

1 · What it is

In an agentic system, the model controls its process and tool use. Once the task is clear, the agent plans and operates independently.

The agent acts and observes the result. It adjusts its approach and repeats until it has completed the task.

Anthropic suggests stopping conditions to maintain control. Agents may return to a human for further information or judgement.

2 · Why it exists

Agentic models coordinate decisions and tool use across several steps.

Choose actionsAn agent uses a language model to manage workflow execution and make decisions.

Use toolsTools extend an agent's capabilities by using application or system APIs.

Adapt the planA self-directed loop can plan, act, observe the result, and adjust what happens next.

3 · How it works

The model directs its process and tool use.

Guardrails are critical for an LLM deployment.

4 · Where it's used

Who	What they ask	What it works with
Operations team	“Can the system complete a multi-step task across connected tools?”	Tool-driven workflow
Agent developer	“Which observation should determine the next action?”	Agent loop state
Risk owner	“Which actions require a human checkpoint?”	Guardrail boundary

5 · What it solves, and what it doesn't

solves

doesn't solve

Anthropic suggests stopping conditions such as a maximum number of iterations.
Sensitive actions can require human oversight.
Anthropic says agents should obtain ground truth from their environment at each step.

6 · Go deeper

This explainer is written in original language. The links below support its factual claims.