Architecture
The system is designed as a hierarchical multi-agent ecosystem. It uses a central Supervisor (Liz) to orchestrate requests across multiple specialized AI Agents, each connected to one or more MCP (Model Context Protocol) servers.
Components
The Supervisor (Liz)
Liz is the entry point of the system and the orchestrator of the user experience. Rather than executing every technical task directly, Liz acts as an intelligent Router.
-
Analyzes the user’s prompt to determine the required domain (e.g., Security, Provisioning, or Fleet).
-
Consults the
AIAgentConfigmetadata to select the most appropriate specialized Agent for the task. -
Tracks the user’s location in the Rancher UI and ensures relevant cluster, namespace, and resource metadata are passed to the downstream Agents to provide a "context-aware" response.
Specialized AI Agents (The "Crew")
The Specialized Agents are the workhorses of the system. Each agent is the orchestrator of its specific domain’s intelligence. It uses a Large Language Model (LLM) as its reasoning engine, while the agent itself provides the structure, coordination, and secure execution of actions.
-
LLM (Reasoning Engine):
-
Interprets user inputs expressed in natural language.
-
Performs the reasoning: breaks down complex requests into smaller steps.
-
Decides what should happen next (keep reasoning vs. take action).
-
Synthesizes outputs from tools into clear, human-readable responses.
-
-
Agent (Orchestrator):
-
Wraps the LLM with the ReAct (Reason + Act) pattern.
-
Decides when to let the LLM keep reasoning or when to act on its instructions.
-
Calls external tools (via the MCP Server) as directed by the LLM.
-
Ensures secure interactions by passing the user’s Rancher token to the MCP server for authentication and authorization.
-
MCP Server
The MCP Server acts as a secure, controlled gateway between the ReAct Agent and the Rancher and Kubernetes APIs. Its functions include:
-
Exposing Tools: It provides a set of well-defined, safe tools (API endpoints) that the ReAct Agent can call. These tools abstract away the complexity of direct Rancher/Kubernetes API interactions.
-
Interaction with Rancher: It translates tool calls from the agent into the appropriate API requests to the Rancher management server, retrieving or modifying resources as needed.
UI Extension
The UI Extension provides the user-facing chat interface within the Rancher dashboard. It is designed to be a seamless part of the Rancher experience and is responsible for:
-
User Input: It captures user queries and sends them to the ReAct Agent.
-
Displaying Responses: It receives responses from the ReAct Agent and presents them in a chat-like format.
How It Works (The Flow)
-
User Request: The user submits a query through the UI Extension.
-
Supervisor Routing: Liz identifies the intent and routes the query, along with the UI context, to the specialized Agent (for example, the Fleet Agent).
-
LLM Reasoning: The specialized Agent’s LLM interprets the request, reasons about the problem, and proposes an action plan.
-
Agent Acting: If the plan requires external operations, the agent calls the appropriate MCP Server tools using the user’s secure token.
-
Response Formulation: The LLM takes the tool outputs and crafts a coherent, human-readable response.
-
Final Response: Liz delivers the Agent’s synthesized answer back to the UI.