Examples

The IBM watsonx Orchestrate ADK repository includes a set of pre-built examples organized by category in the examples/ folder. Use this page to find the example that best matches your use case before diving into the code.

Importing examples

Download the examples

Download the example files to your local computer. You can do this by downloading the entire repository or by copying and creating each file individually.

Activate your environment

Activate your local environment to import the agent and tool examples.

BASH

orchestrate env activate local

Deploy the examples

Deploy the pre-built examples of agents and tools by running the import-all.sh script found in each example folder.

Use the agent and tool examples

Start the watsonx Orchestrate Developer Edition chat and interact with your imported agents and tools.

BASH

orchestrate chat start

Agent builder examples

examples/agent_builder/ End-to-end agent examples built with the ADK. Each subfolder is a self-contained agent that you can import and run.

Example	Description
`agentic_memory`	A healthcare support agent that demonstrates creating a support ticket using a tool, with agentic memory to retain context across turns.
`customer_care`	A customer care agent for a simulated hospital. Queries remote APIs with dummy data about nearby healthcare providers.
`healthcare_provider`	A healthcare provider lookup agent that integrates multiple tools to answer questions about medical facilities and services.
`ibm_knowledge`	An agent backed by an IBM-specific knowledge base. Demonstrates how to wire up a knowledge base with an agent for retrieval-augmented answers.
`personal_banking`	A personal banking agent with Python tools that simulate stateful, long-running operations such as balance inquiries and transaction history.
`voice_enabled_deepgram`	A voice-enabled agent using Deepgram for speech-to-text and text-to-speech.
`voice_enabled_elevenlabs`	A voice-enabled agent using ElevenLabs for high-quality text-to-speech output.
`voice_enabled_google`	A voice-enabled agent using Google Cloud Speech and Text-to-Speech services.
`voice_enabled_watson`	A voice-enabled agent using IBM Watson Speech to Text and Text to Speech services.

Agentic workflow (flow builder) examples

examples/flow_builder/ Examples built using the watsonx Orchestrate Agentic Workflow (flow builder). Each subfolder contains a flow definition and any supporting agents or tools needed to run it.

Example	Description
`Document_Review`	A workflow that routes documents through a review process, demonstrating approval steps and conditional branching.
`Employee_information`	A flow for looking up and presenting employee information from an internal data source.
`Event_manager`	A workflow that manages event scheduling and notifications across participants.
`Grievance_Processing_Agent`	A multi-step grievance processing flow that triages, routes, and tracks employee complaints.
`HR_agent`	An HR assistant workflow covering common HR requests such as time-off and policy lookups.
`IT_agent`	An IT helpdesk flow that handles support ticket creation and routing.
`Payment_agent`	A payment processing workflow demonstrating approval gates, validation steps, and error handling.
`Random_Failure`	A testing flow that intentionally introduces random failures to demonstrate error-handling and retry patterns.
`Time_off_Agent`	A time-off request workflow with manager approval and calendar integration.
`agent_scheduler`	Demonstrates scheduling agent invocations at a defined cadence using the flow scheduler.
`book_a_flight`	A flight booking workflow that collects traveler details, searches for flights, and confirms a booking.
`car_price_calculator`	A flow that calculates car prices based on user-provided options and configuration inputs.
`collaborator_agents`	Demonstrates how to wire multiple collaborator agents together within a single workflow.
`document_classifier`	A workflow that classifies uploaded documents using the ADK document classifier and routes them accordingly.
`document_extractor`	Extracts key-value data from documents using the ADK text extractor.
`document_extractor_structured`	An enhanced document extractor that uses a structured schema, including table fields and `available_options`.
`document_processing`	An end-to-end document processing pipeline combining classification, extraction, and downstream routing.
`dynamic_forms`	Demonstrates how to build flows that render dynamic forms to collect structured user input.
`extract_support_request`	Extracts structured fields from a free-text support request and creates a ticket.
`flow_callback`	Demonstrates the callback pattern for flows that need to pause and resume upon an external event, with database persistence.
`flow_callback_no_db`	A simplified callback pattern without database persistence — useful for stateless callback scenarios.
`flow_mcp_tester`	A testing flow for MCP (Model Context Protocol) tool integrations within a workflow.
`foreach_email`	Iterates over a list of recipients and sends a personalized email to each, demonstrating the `forEach` loop construct.
`get_facts_numbers`	A simple flow that retrieves facts and numeric data — a good starting point for understanding basic flow structure.
`get_insurance_rate`	A flow that calculates insurance rates based on user inputs and business rules.
`get_pet_facts_error_branching`	Retrieves pet facts while demonstrating error-handling branches and fallback behavior.
`get_pet_facts_if_else`	Retrieves pet facts using if/else conditional branching — a beginner-friendly example for flow control.
`get_tuition_reimbursed`	A tuition reimbursement request flow with validation, approval, and notification steps.
`hello_message_flow`	The simplest possible flow — sends a hello message. Start here if you are new to the flow builder.
`hello_message_flow_datamap`	Extends the hello flow to demonstrate data mapping between flow steps.
`hello_message_script_flow`	Extends the hello flow to show how to embed custom scripts within a flow step.
`ibm_knowledge_to_emails`	Queries an IBM knowledge base and sends the results as formatted emails.
`live_agent_transfer_flow`	Demonstrates how to transfer a conversation from an AI agent to a human live agent within a flow.
`masking_test_flow`	Tests data masking behavior for sensitive fields within a flow.
`parallel_flow`	Runs multiple flow branches in parallel and merges results, demonstrating concurrent execution.
`recent_transactions`	Fetches and presents a user’s recent transaction history from a simulated banking data source.
`schedule_helpdesk_alert`	Schedules and sends helpdesk alert notifications based on a defined trigger.
`text_extraction`	Extracts plain text from an uploaded document and passes it to downstream flow steps.
`triage_workflow_agent_swarm`	A triage workflow using a swarm of collaborator agents to classify and route requests.
`user_activity`	Tracks and reports on user activity, with file outputs and user assignment.
`user_activity_no_files`	A simplified variant of `user_activity` without file output steps.
`user_activity_user_assignment`	Extends `user_activity` to demonstrate dynamic user assignment within a flow.
`user_activity_with_forms`	Extends `user_activity` with inline form steps for collecting additional user input mid-flow.

Channel integration examples

examples/channel-integrations/ Examples showing how to connect an agent to a specific messaging or telephony channel.

Example	Description
`SIP_phone_channel`	Connects an agent to a SIP trunk phone channel for inbound/outbound voice calls.
`facebook_messenger_channel`	Integrates an agent with Facebook Messenger for conversational messaging.
`genesys_audio_connector_phone_channel`	Connects an agent to a Genesys Cloud contact center via the Audio Connector for voice interactions.
`genesys_bot_connector_channel`	Integrates an agent as a bot within Genesys Cloud using the Bot Connector API.
`slack_channel`	Deploys an agent as a Slack app for team-based conversational interactions.
`teams_channel`	Deploys an agent as a Microsoft Teams bot.
`twilio_sms_channel`	Connects an agent to Twilio for SMS-based interactions.
`twilio_whatsapp_channel`	Connects an agent to Twilio for WhatsApp messaging.

Evaluation examples

examples/evaluations/ Examples for testing, evaluating, and benchmarking agents using the ADK evaluation framework.

Example	Description
`evaluate`	Core evaluation flow: run test cases against an agent and score the results.
`evaluate_with_context_variable`	Demonstrates running evaluations that include context variables passed to the agent at evaluation time.
`evaluate_with_file_upload`	Evaluates agents that accept file inputs, testing file upload handling as part of the interaction.
`generate`	Automatically generates test cases from user stories and tool definitions using the `generate` command.
`analysis`	Analyzes evaluation results and produces summary reports for pass/fail rates and scoring.
`native_agent_validation`	Validates a native ADK agent against a predefined set of test cases.
`external_agent_validation`	Validates an external agent (non-native) integrated with watsonx Orchestrate.
`quick-eval`	A fast, minimal evaluation setup for rapid iteration and smoke testing during development.
`red-teaming`	Runs adversarial test inputs against an agent to surface safety, robustness, and guardrail issues.
`rubric_evals`	Uses rubric-based scoring to evaluate agent responses against defined criteria.
`environment_manager`	Helper utilities for managing evaluation environments and configuration.
`cpd`	Evaluation examples targeting IBM Cloud Pak for Data (CPD) deployments.

Custom agents examples

examples/custom_agents/ Examples of external agents built using the watsonx Orchestrate SDK (Python).

Example	Description
`local_memory_sdk_agent`	An SDK-based agent that maintains conversation memory in local storage, demonstrating stateful agent patterns without an external memory service.
`memory_coach_sdk_agent`	A coaching agent that uses memory to recall prior session context and provide personalized guidance across conversations.

Langflow examples

examples/langflow/ Examples of agents built with Langflow and connected to watsonx Orchestrate.

Example	Description
`travel_advice`	A Langflow-based travel advisor agent that answers questions about destinations and travel planning.
`travel_visa_enquiry`	A Langflow-based agent that handles travel visa enquiries, demonstrating document and policy lookup.

Plugins examples

examples/plugins/ A self-contained example demonstrating how to build and register OpenAPI-based plugins (tools) and wire them to an agent. The folder is organized into three parts that work together:

Subfolder	Description
`agents/`	Agent definition that uses the plugins as tools.
`plugins/`	OpenAPI plugin definitions imported as tools.
`tools/`	Python tool implementations backing the plugin endpoints.

Observability examples

examples/observability/

Example	Description
`traces/`	Demonstrates how to export and visualize agent execution traces for debugging and monitoring. Includes a detailed README with setup instructions.

Voice runtime example

examples/voice/ A browser-based testing tool for voice agents. Provides real-time conversation testing, call recording, and CDR (Call Detail Record) monitoring. Run it locally with uvicorn app:app and open your browser to interact with your voice agent.

Standalone example

examples/hotel_list_agent/ A standalone agent example that lists hotels using a Python tool with hardcoded dummy data. Useful as a quick reference for the minimal structure required to define and import an agent.

Experimental examples

examples/experimental/ Preview examples for features that are still under development. These may change or be removed without notice.

Example	Description
`agent_builder/customer_care_style`	An experimental variant of the customer care agent exploring a new agent style configuration.
`flow_builder/flow_converter`	An experimental utility for converting flow definitions between formats.

Release Notes

Get Started

Build

Deploy

Analyze

Developer experience

Legal notices

Importing examples

Agent builder examples

Agentic workflow (flow builder) examples

Channel integration examples

Evaluation examples

Custom agents examples

Langflow examples

Plugins examples

Observability examples

Voice runtime example

Standalone example

Experimental examples

​Importing examples

​Agent builder examples

​Agentic workflow (flow builder) examples

​Channel integration examples

​Evaluation examples

​Custom agents examples

​Langflow examples

​Plugins examples

​Observability examples

​Voice runtime example

​Standalone example

​Experimental examples

Importing examples

Agent builder examples

Agentic workflow (flow builder) examples

Channel integration examples

Evaluation examples

Custom agents examples

Langflow examples

Plugins examples

Observability examples

Voice runtime example

Standalone example

Experimental examples