Workflows and Multi-Agent

A normal run is one goal plus tools. Sometimes you need more: several steps after each other, or several specialized agents. For this, AiCleverness has workflows and a router agent.

Defining a Workflow

A workflow is a list of connected steps (nodes):

// A workflow where one step runs after the other
var workflow = new WorkflowDefinition(
    Name: "research-pipeline",
    Nodes: [
        new WorkflowNode("search", "tool-execution", new() { ["tool"] = "search_web" }),
        new WorkflowNode("analyze", "agent-execution", new() { ["goal"] = "Analyze results" }),
        new WorkflowNode("report", "agent-execution", new() { ["goal"] = "Write report" })
    ]);

Each node has a name, a type, and parameters. There are two node types:

• tool-execution — the node runs one registered tool.
• agent-execution — the node runs a complete agent run with its own goal.

WorkflowResult collects the output of every node.

Execution

services.AddWorkflowExecutor<SequentialWorkflowExecutor>();

IWorkflowExecutor is the interface. The library contains a sequential executor (one node after the other). If your nodes have dependencies (node B needs the result of node A), you can express this in the node graph and write your own executor.

Router Agent

If you have several specialized agents, one router can decide which one gets the request:

services.AddRouterAgent<MyRouterAgent>();

Implement IRouterAgent: it looks at the request and sends it to the right named agent. Together with agent scoping, every agent keeps its own policies, gates, and validators.