Agents don’t always need the same tools at every step. With dynamic tool calling, you can now control which tools are available at different points in a run.

Why it matters:

• Enforce workflows: Require an auth tool before exposing sensitive actions.

• Guide reasoning: Start with a small, focused toolset, then expand as the task evolves.

• Improve reliability: Reduce wrong turns by limiting unnecessary tools early.

Think of it as giving your agent the right toolbox at the right time—so it’s safer, more efficient, and easier to control.

Read the docs: https://langchain-ai.github.io/langgraph/how-tos/tool-calling/?h=dynamic#use-in-an-agent