Sequential Agents

The SequentialAgent

Sequential agents execute sub‑agents one after another in a predetermined order, creating predictable pipeline workflows where each step builds on the previous one. Use the SequentialAgent when you need a fixed, strict order (for example: get → analyze → summarize). As with other workflow agents, orchestration is deterministic; sub‑agents can be any type, including LlmAgent.

Visual

Quick Example

Here's how to create a research pipeline that processes information through multiple sequential steps:

import { LlmAgent, SequentialAgent } from "@iqai/adk";

// Create specialized agents for each step
const researchAgent = new LlmAgent({
  name: "researcher",
  description: "Gathers information",
  model: "gemini-2.5-flash",
  instruction: "Gather comprehensive information about the topic",
  // tools: [new GoogleSearchTool(), new WikipediaTool()],
});

const analysisAgent = new LlmAgent({
  name: "analyst",
  description: "Analyzes research data",
  model: "gemini-2.5-flash",
  instruction: "Analyze the research data and identify key insights",
});

const summaryAgent = new LlmAgent({
  name: "summarizer",
  description: "Summarizes findings",
  model: "gemini-2.5-flash",
  instruction: "Create a concise summary with actionable recommendations",
});

// Create sequential pipeline that runs research → analysis → summary
const pipeline = new SequentialAgent({
  name: "research_pipeline",
  description: "Research, analysis, summary",
  subAgents: [researchAgent, analysisAgent, summaryAgent],
});

How it Works

When a SequentialAgent runs:

1. Iteration: It iterates through subAgents in the order provided
2. Sub‑Agent Execution: For each sub‑agent, it calls the sub‑agent’s run method

Key Benefits

• Ordered Processing: Each step builds on the previous agent's output
• Predictable Flow: Deterministic execution order for reliable workflows
• Context Continuity: Information flows naturally through the conversation
• Specialized Roles: Each agent can focus on its specific expertise

When to Use

Ideal for:

• Multi-step pipelines (research → analysis → summary)
• Data processing workflows (extract → transform → load)
• Content creation (draft → review → finalize)
• Quality assurance (generate → validate → approve)

Avoid when:

• Tasks are independent and can run simultaneously
• Order doesn't matter for the final result
• You need faster parallel processing

Related Topics