Skip to content

LangChain Agent Architecture

A traditional chatbot maps each user message to a fixed intent and a fixed handler. A LangChain agent works differently: an LLM reasons about the query, decides which tool to call, runs it, reads the result, and may call another tool before answering. This MicroSim animates that loop so you can watch the agent choose its own path through five available tools.

Interactive Demo

Run MicroSim Fullscreen

To embed this MicroSim in your own page, use the following iframe:

1
<iframe src="main.html" width="100%" height="662" scrolling="no"></iframe>

Overview

Pick an example query from the dropdown and press Run Agent. The animation highlights each stage of the agent loop:

  1. The User Query box lights up (blue).
  2. The LLM Reasoning box (green) shows the agent's thought about which tool it needs.
  3. One of the five tools (orange) is selected and executes: Vector DB Search, SQL Query, Calculator, Web Search, or Custom API.
  4. The tool result feeds back to the LLM (dashed green loop).
  5. For multi-step queries the loop repeats with a second tool.
  6. The Final Response box (blue) shows the answer.

The Execution Log on the right records every step with a simulated timestamp.

Controls:

  • Example query dropdown - choose a single-tool or multi-tool scenario.
  • Run Agent / Reset buttons.
  • Animation speed slider - 100 to 2000 ms per step.
  • Show intermediate reasoning checkbox - include or hide the LLM's tool-selection thoughts in the log.

Lesson Plan

  • Contrast with intent routing. After running a query, ask students how a fixed intent-to-handler system would have struggled with the multi-step calculator example.
  • Predict the tool. Pause before pressing Run and have students predict which tool the agent will pick and why.
  • Trace the loop. Run the multi-step query and have students count how many times control returns to the LLM box.
  • Toggle reasoning. Turn off "Show intermediate reasoning" and discuss why hiding the agent's chain-of-thought makes debugging harder.

References