All Classes and Interfaces
Class
Description
Base class for MCP tools.
Manages streaming tool related resources during invocation.
Provides ADK components that are part of core.
Service provider interface for ADK components to be registered in
ComponentRegistry.AgentTool implements a tool that allows an agent to call another agent.
RequestProcessor that handles agent transfer for LLM flow.Annotations for tools.
The annotation for binding the 'Schema' input.
A
BaseLlm implementation for calling an Apigee proxy.Builder for
ApigeeLlm.The API response contains a response to a call to the GenAI APIs.
Represents an LLM-backed agentic application.
Builder for
App.Application Integration Toolset
LLM flow with automatic agent transfer support.
Base class for all agents.
Base Builder for all agents.
Base configuration for all agents with subagent support.
Configuration for referencing other agents (subagents).
Reference to a callback stored in the ComponentRegistry.
Base interface for artifact services.
Abstract base class for all code executors.
Defines an interface for computer environments.
Base interface for producing events summary.
An interface that provides examples for a given query.
Interface for the execution flows to run a group of agents.
Abstract base class for Large Language Models (LLMs).
The base class for a live model connection.
A basic flow that calls the LLM in a loop until a final response is generated.
Base contract for memory services.
Base class for creating plugins.
Base class for retrieval tools.
The base class for all ADK tools.
Configuration class for tool arguments that allows arbitrary key-value pairs.
Configuration class for a tool definition in YAML/JSON.
Base interface for toolsets.
RequestProcessor that handles basic information to build the LLM request.A built-in code execution tool that is automatically invoked by Gemini 2 models.
A code executor that uses the Model's built-in code executor.
The context of various callbacks for an agent invocation.
Functional interfaces for agent lifecycle callbacks.
Async callback interface for actions to be performed after an agent has finished running.
Helper interface to allow for sync afterAgentCallback.
Helper interface to allow for sync afterModelCallback.
Async callback interface for actions to be performed after a tool has been invoked.
Helper interface to allow for sync afterToolCallback.
Async callback interface for actions to be performed before an agent starts running.
Helper interface to allow for sync beforeAgentCallback.
Helper interface to allow for sync beforeModelCallback.
Async callback interface for actions to be performed before a tool is invoked.
Helper interface to allow for sync beforeToolCallback.
Async callback interface for handling errors that occur during an LLM model call.
Helper interface to allow for sync onModelErrorCallback.
Async callback interface for handling errors that occur during a tool invocation.
Helper interface to allow for sync onToolErrorCallback.
Utility methods for normalizing agent callbacks.
Represents the Claude Generative AI model by Anthropic.
Handles Code Execution related logic.
Utility functions for code execution.
A structure that contains the input of code execution.
Builder for
CodeExecutionUtils.CodeExecutionInput.A structure that contains the result of code execution.
Builder for
CodeExecutionUtils.CodeExecutionResult.A structure that contains a file name and its content.
Builder for
CodeExecutionUtils.File.The persistent context used to configure the code executor.
Frequently used code snippets for collections.
Request processor that performs event compaction.
A registry for storing and retrieving ADK instances by name.
Enum for computer environments.
Represents the current state of the computer environment.
Builder for
ComputerState.A tool that wraps computer control functions for use with LLMs.
A toolset that provides computer use capabilities.
Utility class for loading agent configurations from YAML files.
Exception thrown when configuration is invalid.
Utility class for interacting with the Google Cloud Connectors API.
Represents the schema for an action.
Represents details of a connection.
Represents the schema and available operations for an entity.
A code executor that uses a custom container to execute code.
RequestProcessor that populates content in request for LLM flows.Configuration for context caching across all agents in an app.
A plugin that filters the LLM request
Content list to reduce its size, for example to
adhere to context window limits.Builder for
ContextFilterPlugin.Utility class for converting between different representations of MCP tools.
Provides ADK components that are part of core.
This interface provides a method to convert a service account JSON string to a Google Credentials
object.
The default builder for creating MCP client transports.
Represents an event in a session.
Builder for
Event.Represents the actions attached to an event.
Builder for
EventActions.The compaction of the events.
Builder for
EventCompaction.Base interface for compacting events.
Configuration for event compaction.
Builder for
EventsCompactionConfig.Iterable stream of
Event objects.Represents an few-shot example.
Builder for constructing
Example instances.RequestProcessor that populates examples in LLM request.A tool that injects (few-shot) examples into the outgoing LLM request as system instructions.
Builder for
ExampleTool.Utility class for examples.
Tool for exiting execution of
LoopAgent.Utility class for function calling.
Utility class for handling function calls.
FunctionTool implements a customized function calling tool.
An artifact service implementation using Google Cloud Storage (GCS).
Represents the Gemini Generative AI model.
Builder for
Gemini.Manages a persistent, bidirectional connection to the Gemini model via WebSockets for real-time
interaction.
Request / Response utilities for
Gemini.Configuration for getting a session.
Builder for
GetSessionConfig.Plugin that provides global instructions functionality at the App level.
A built-in tool that is automatically invoked by Gemini 2 models to retrieve search results from
Google Maps.
A tool that wraps a sub-agent that only uses google_search tool.
A built-in tool that is automatically invoked by Gemini 2 and 3 models to retrieve search results
from Google Search.
Base client for the HTTP APIs.
Wraps a real HTTP response to expose the methods needed by the GenAI SDK.
RequestProcessor that gives the agent identity from the frameworkAn in-memory implementation of the
BaseArtifactService.An in-memory memory service for prototyping purposes only.
The class for the in-memory GenAi runner, using in-memory artifact and session services.
An in-memory implementation of
BaseSessionService assuming Session objects are
mutable regarding their state map, events list, and last update time.Represents an instruction that can be provided to an agent to guide its behavior.
Returns an instruction dynamically constructed from the given context.
Plain instruction directly provided to the agent.
RequestProcessor that handles instructions and global instructions for LLM flows.Utility methods for handling instruction templates.
Utility class for interacting with Google Cloud Application Integration.
Application Integration Tool
The context for an agent invocation.
Builder for
InvocationContext.The base class for the types that needs JSON serialization/deserialization capability.
Response for listing artifacts.
Builder for
ListArtifactsResponse.Response for listing artifact versions.
Builder for
ListArtifactVersionsResponse.Response for listing events.
Builder for
ListEventsResponse.Response for listing sessions.
Builder for
ListSessionsResponse.Represents a request to be sent to a live connection to the LLM model.
Builder for constructing
LiveRequest instances.A queue of live requests to be sent to the model.
The LLM-based agent.
Builder for
LlmAgent.Enum to define if contents of previous events should be included in requests to the underlying
LLM.
Configuration for LlmAgent.
An error indicating that the limit for calls to the LLM has been exceeded.
An LLM-based event summarizer for sliding window compaction.
Central registry for managing Large Language Model (LLM) instances.
The factory interface for creating LLM instances.
Represents a request to be sent to the LLM.
Builder for constructing
LlmRequest instances.Represents a response received from the LLM.
Builder for constructing
LlmResponse instances.A tool that loads artifacts and adds them to the session.
The response from a load memory tool invocation.
A tool that loads memory for the current user.
A plugin that logs important information at each callback point.
A function tool that returns the result asynchronously.
An agent that runs its sub-agents sequentially in a loop.
Builder for
LoopAgent.Configuration for LoopAgent.
Initializes a MCP tool.
Connects to a MCP Server, and retrieves MCP Tools into ADK Tools.
Builder for McpAsyncToolset
Manages MCP client sessions.
Initializes a MCP tool.
Base exception for all errors originating from
AbstractMcpTool and its subclasses.Exception thrown when there's an error during MCP tool declaration generated.
Connects to a MCP Server, and retrieves MCP Tools into ADK Tools.
Configuration class for MCPToolset.
Base exception for all errors originating from
McpToolset.Exception thrown when there's an error during MCP session initialization.
Exception thrown when there's an error during loading tools from the MCP server.
Interface for building McpClientTransport instances.
Represents one memory entry.
Builder for
MemoryEntry.Represents a model by name or instance.
Builder for
Model.Utility class for creating ConcurrentHashMaps.
A shell agent that runs its sub-agents in parallel in isolated manner.
Builder for
ParallelAgent.Configuration for ParallelAgent.
Interface for creating plugins.
Manages the registration and execution of plugins.
Provides read-only access to the context of an agent run.
Handles tool confirmation information to build the LLM request.
Interface for processing LLM requests.
Result of request processing.
Interface for processing LLM responses.
Result of response processing.
Configuration to modify an agent's LLM's underlying behavior.
Builder for
RunConfig.Streaming mode for the runner.
Tool execution mode for the runner, when they are multiple tools requested (by the models or
callbacks).
The main class for the GenAI Agents runner.
Builder for
Runner.Utility class for validating schemas.
Represents the response from a memory search.
Builder for
SearchMemoryResponse.An agent that runs its sub-agents sequentially.
Builder for
SequentialAgent.Configuration for SequentialAgent.
Builder for
Session.Represents a general error that occurred during session management operations.
Key for a session, composed of appName, userId and session id.
Indicates that a requested session could not be found.
Utility functions for session service.
Basic LLM flow with fixed request and response processors.
This class performs events compaction in a sliding window fashion based on the
EventsCompactionConfig.Interface for a speech-to-text client.
Parameters for establishing a MCP Server-Sent Events (SSE) connection.
Builder for
SseServerParameters.A
State object that also keeps track of the changes to the state.Parameters for establishing a MCP stdio connection.
Builder for
StdioServerParameters.Server parameters for Streamable HTTP client transport.
Builder for
StreamableHttpServerParameters.This class performs event compaction by retaining the tail of the event stream.
Represents a tool confirmation configuration.
Builder for
ToolConfirmation.ToolContext object provides a structured context for executing tools or functions.
Builder for
ToolContext.Functional interface to decide whether a tool should be exposed to the LLM based on the current
context.
Utility class for capturing and reporting telemetry data within the ADK.
A transformer that re-activates a given context for the duration of the stream's subscription.
A transformer that manages an OpenTelemetry span and scope for RxJava streams.
A built-in tool that is automatically invoked by Gemini 2 and 3 models to retrieve information
from the given URLs.
Tracks the current ADK version.
A code executor that uses Vertex Code Interpreter Extension to execute code.
A retrieval tool that fetches context from Vertex AI RAG.
A tool that wraps a sub-agent that only uses vertex_ai_search tool.
A built-in tool using Vertex AI Search.
Builder for
VertexAiSearchTool.Connects to the managed Vertex AI Session Service.
Credentials for accessing Gemini models through Vertex.
Builder for
VertexCredentials.Implementation of SpeechClientInterface using Vertex AI SpeechClient.