A supervisor/orchestrator agent that listens to user chat, routes the conversation to the appropriate tool, and then calls the chosen agent.
This agent is a building-block base pattern agent that takes user chat conversations and initiates any approprioate sub-agents.
It starts with a user chat conversation.
The 'Agent Router' examines the conversation history using an LLM call. It analyzes both the previous messages and the latest user input to decide which specialized tool should handle the request: 'Agent 1', 'Agent N', or, if no agent is a good match, 'General User Input'.
When one of the sub-agents (such as 'Agent 1' or 'Agent N') is called, it executes its domain-specific processing with its capabilities. Each sub-agent focuses on handling a particular category of queries, ensuring that the response is precise and efficiently processed.
For cases where the input does not clearly indicate a specialized need that the sub-agents can handle, the 'general user input' is utilized. This handles general questions and provides a response that may also include a brief overview of available tools.
After the selected sub-agent has processed the request, the workflow concludes.
Get ideas for agent-powered experiences that help prospects, users, customers, and partners discover, evaluate, and get value from your product.
Find new ways to improve the internal workflows your team uses to launch, position, sell, support, and grow those experiences across marketing, sales, growth, and success.
See examples of agent experiences you can build for prospects, users, customers, and partners.
Discover team workflows you can turn into internal agents, automations, and tools that save time and reduce manual work.
Find ideas for better experiences during the of the buyer or customer journey.
Explore internal workflows that help your team support that stage more effectively.
Use this example as a starting point, not a fixed recipe.
Copy the flow above as a prompt, markdown, skill, or mermaid diagrom. Adjust the steps, tools, agent and human involvement to fit your real workflow. Then test where users actually get value or drop off.
You are a routing assistant. Analyze the current conversation and choose one of the following routes based on the user's query: agent_1, agent_n, or general_user_input.
You are an AI assistant. If the user asks what you can do, describe the available tools. Otherwise, respond directly to the query in a clear and friendly manner.