What are all the different types of agents in Google ADK?

Types of Agents in Google ADK

Google ADK supports several agent types, each suited for different use cases in multi-agent architectures.

1. LLM Agent (Base Agent)

The fundamental building block — an agent powered by a language model:

pythonCopy
from google.adk.agents import LlmAgent
from google.adk.tools import google_search

research_agent = LlmAgent(
    name="researcher",
    model="gemini-2.0-flash",
    description="Searches the web for information",
    instruction="You are a research specialist. Search and summarize accurately.",
    tools=[google_search]
)

Use when: You need an agent that can reason, use tools, and respond in natural language.

2. Sequential Agent

Executes a fixed sequence of sub-agents in order:

pythonCopy
from google.adk.agents import SequentialAgent

pipeline = SequentialAgent(
    name="research_pipeline",
    description="Research then write then review",
    sub_agents=[
        search_agent,    # Step 1: gather info
        writer_agent,    # Step 2: write content
        reviewer_agent   # Step 3: review output
    ]
)

Use when: Tasks have a clear, fixed order of steps.

3. Parallel Agent

Runs multiple sub-agents simultaneously:

pythonCopy
from google.adk.agents import ParallelAgent

parallel_researcher = ParallelAgent(
    name="multi_source_research",
    description="Research from multiple sources simultaneously",
    sub_agents=[
        web_search_agent,     # Runs simultaneously
        database_agent,       # Runs simultaneously
        document_agent        # Runs simultaneously
    ]
)
# All three run at once; results are combined

Use when: Subtasks are independent and can be parallelized for speed.

4. Loop Agent

Repeatedly executes sub-agents until a condition is met:

pythonCopy
from google.adk.agents import LoopAgent

improvement_loop = LoopAgent(
    name="iterative_improver",
    description="Improve content until quality threshold is met",
    sub_agents=[writer_agent, quality_checker_agent],
    max_iterations=5
)

Use when: Tasks need iterative refinement (e.g., write → review → rewrite).

5. Custom Agent (BaseAgent)

Full control over agent behavior:

pythonCopy
from google.adk.agents import BaseAgent
from google.adk.agents.invocation_context import InvocationContext
from google.adk.events import Event
from typing import AsyncGenerator

class SmartRoutingAgent(BaseAgent):
    '''Custom agent with domain-based routing'''

    async def _run_async_impl(
        self, ctx: InvocationContext
    ) -> AsyncGenerator[Event, None]:
        user_message = ctx.user_content.parts[0].text

        # Custom routing logic
        if "data" in user_message.lower():
            async for event in self.data_agent.run_async(ctx):
                yield event
        elif "code" in user_message.lower():
            async for event in self.code_agent.run_async(ctx):
                yield event
        else:
            async for event in self.general_agent.run_async(ctx):
                yield event

Use when: You need routing logic, conditional execution, or custom orchestration patterns.

Agent Type Selection Guide

Hierarchy Example

pythonCopy
# Full multi-agent hierarchy
orchestrator = LlmAgent(
    name="coordinator",
    model="gemini-2.0-flash",
    sub_agents=[
        SequentialAgent("research_pipeline", sub_agents=[searcher, analyzer]),
        ParallelAgent("multi_search", sub_agents=[web_agent, db_agent]),
        LoopAgent("refiner", sub_agents=[writer, reviewer])
    ]
)