google-adk-kotlin-core/com.google.adk.kt.agents/LlmAgent

LlmAgent

class LlmAgent(val name: String, val model: Model, val description: String = "", val subAgents: List<BaseAgent> = emptyList(), val beforeAgentCallbacks: List<BeforeAgentCallback> = emptyList(), val afterAgentCallbacks: List<AfterAgentCallback> = emptyList(), val disallowTransferToParent: Boolean = false, val disallowTransferToPeers: Boolean = false, val tools: List<BaseTool> = emptyList(), val toolsets: List<Toolset> = emptyList(), val generateContentConfig: GenerateContentConfig? = null, val instruction: Instruction? = null, val staticInstruction: Content? = null, val beforeModelCallbacks: List<BeforeModelCallback> = emptyList(), val afterModelCallbacks: List<AfterModelCallback> = emptyList(), val beforeToolCallbacks: List<BeforeToolCallback> = emptyList(), val afterToolCallbacks: List<AfterToolCallback> = emptyList(), val inputSchema: Schema? = null, val onModelErrorCallbacks: List<OnModelErrorCallback> = emptyList(), val onToolErrorCallbacks: List<OnToolErrorCallback> = emptyList(), val includeContents: LlmAgent.IncludeContents = IncludeContents.DEFAULT) : BaseAgent

LLM-based Agent.

When this agent is a sub-agent and the parent transfers control to it via transfer_to_agent, the runner decides who handles the next user turn based on the disallowTransferToParent / disallowTransferToPeers flags inherited from BaseAgent - see those flags for the full dispatch rules.

Constructors

LlmAgent

constructor(name: String, model: Model, description: String = "", subAgents: List<BaseAgent> = emptyList(), beforeAgentCallbacks: List<BeforeAgentCallback> = emptyList(), afterAgentCallbacks: List<AfterAgentCallback> = emptyList(), disallowTransferToParent: Boolean = false, disallowTransferToPeers: Boolean = false, tools: List<BaseTool> = emptyList(), toolsets: List<Toolset> = emptyList(), generateContentConfig: GenerateContentConfig? = null, instruction: Instruction? = null, staticInstruction: Content? = null, beforeModelCallbacks: List<BeforeModelCallback> = emptyList(), afterModelCallbacks: List<AfterModelCallback> = emptyList(), beforeToolCallbacks: List<BeforeToolCallback> = emptyList(), afterToolCallbacks: List<AfterToolCallback> = emptyList(), inputSchema: Schema? = null, onModelErrorCallbacks: List<OnModelErrorCallback> = emptyList(), onToolErrorCallbacks: List<OnToolErrorCallback> = emptyList(), includeContents: LlmAgent.IncludeContents = IncludeContents.DEFAULT)

Types

IncludeContents

enum IncludeContents : Enum<LlmAgent.IncludeContents>

Controls how prior conversation history is included in this agent's model request.

Properties

afterAgentCallbacks

val afterAgentCallbacks: List<AfterAgentCallback>

List of callbacks to run after the agent executes.

afterModelCallbacks

val afterModelCallbacks: List<AfterModelCallback>

List of callbacks to run after each model call.

afterToolCallbacks

val afterToolCallbacks: List<AfterToolCallback>

List of callbacks to run after each tool call.

beforeAgentCallbacks

val beforeAgentCallbacks: List<BeforeAgentCallback>

List of callbacks to run before the agent executes.

beforeModelCallbacks

val beforeModelCallbacks: List<BeforeModelCallback>

List of callbacks to run before each model call.

beforeToolCallbacks

val beforeToolCallbacks: List<BeforeToolCallback>

List of callbacks to run before each tool call.

description

open val description: String

The description of the agent.

disallowTransferToParent

val disallowTransferToParent: Boolean = false

When true, the framework will not route the next user turn back to this agent after the parent transfers control to it; instead the next turn falls back to the root agent. Set this on utility sub-agents the parent calls and returns from (translators, summarizers, classifiers). Leave at the default false for sub-agents that should keep handling follow-up turns directly (e.g. billing, support).

disallowTransferToPeers

val disallowTransferToPeers: Boolean = false

When true, prevents this agent from transferring sideways to a peer agent under the same parent. Typically set together with disallowTransferToParent on one-shot utility agents. Violations are surfaced by the runner as IllegalArgumentException.

generateContentConfig

val generateContentConfig: GenerateContentConfig? = null

The additional content generation configurations.

includeContents

val includeContents: LlmAgent.IncludeContents

Controls how prior conversation history is included in the model request. Defaults to IncludeContents.DEFAULT, which includes the relevant conversation history. Set to IncludeContents.NONE to exclude prior history; the model then receives only the current turn (the most recent user input or other-agent reply, plus any tool calls/responses produced within that turn). The system instruction and tools are preserved in both modes.

inputSchema

val inputSchema: Schema? = null

The input schema of the agent.

instruction

val instruction: Instruction? = null

Instruction guiding the agent's behavior. Use one of: - Instruction("text") for a literal string (the most common case), - Instruction(content) for a pre-built, possibly multimodal Content, or - Instruction { ctx -> ... } for a Instruction.Provider resolved per turn.

model

val model: Model

The model to use for the agent.

name

val name: String

The name of the agent.

onModelErrorCallbacks

val onModelErrorCallbacks: List<OnModelErrorCallback>

List of callbacks to run when a model call fails.

onToolErrorCallbacks

val onToolErrorCallbacks: List<OnToolErrorCallback>

List of callbacks to run when a tool call fails.

staticInstruction

val staticInstruction: Content? = null

Static instruction content sent literally as system instruction at the beginning. This field is for content that never changes. It's sent directly to the model without any processing or variable substitution.

subAgents

val subAgents: List<BaseAgent>

List of sub-agents.

tools

val tools: List<BaseTool>

Tools available to this agent.

toolsets

val toolsets: List<Toolset>

Toolsets available to this agent.

Functions

findAgent

fun BaseAgent.findAgent(targetName: String): BaseAgent?

Finds an agent with the given name in this agent's subtree (including itself).

runAsync

fun runAsync(parentContext: InvocationContext): Flow<Event>

Public entry point for executing the agent asynchronously (text-based).