ADK Python

https://github.com/google/adk-python
Admin
An open-source, code-first Python toolkit for building, evaluating, and deploying sophisticated AI...
Tokens:485,499
Snippets:3,885
Trust Score:10
License:Apache-2.0
Update:1 week ago
Show doc for...
Tokens:
Raw
### Run single-process LiveKit quickstart

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/livekit/README.md

Installs prerequisites and launches the single-process LiveKit quickstart sample. Access the client at http://localhost:8080 to test voice and camera interactions.

```bash
pip install "google-adk[livekit]" fastapi uvicorn
python -m contributing.samples.integrations.livekit.quickstart
```

--------------------------------

### Run with Dev UI

Source: https://github.com/google/adk-python/blob/main/contributing/samples/models/hello_world_openai/README.md

Start the Dev UI to interact with the agent. Open the printed URL, select hello_world_openai, and try the example prompt.

```bash
uv run --extra extensions adk web contributing/samples/models/hello_world_openai
```

--------------------------------

### Create an ExampleTool with Example objects and attach it to an Agent

Source: https://github.com/google/adk-python/blob/main/docs/guides/examples/example/index.md

Build Example objects with UserContent input and ModelContent outputs, then pass the ExampleTool in the agent's tools list. The examples are not a callable tool; they arrive as system instruction text on every turn alongside the instruction.

```python
from google.adk.agents import Agent
from google.adk.examples import Example
from google.adk.tools import ExampleTool
from google.genai import types

example_tool = ExampleTool([
    Example(
        input=types.UserContent(parts=[types.Part(text="Where is order 4417?")]),
        output=[
            types.ModelContent(
                parts=[types.Part.from_function_call(
                    name="check_order_status", args={"order_id": "4417"}
                )]
            ),
            types.ModelContent(
                parts=[types.Part(text="Order 4417 shipped on Tuesday and arrives Friday.")]
            ),
        ],
    ),
    Example(
        input=types.UserContent(parts=[types.Part(text="I want a refund.")]),
        output=[
            types.ModelContent(
                parts=[types.Part(text="Sure — which order number is that for?")]
            )
        ],
    ),
])

agent = Agent(
    name="support_agent",
    instruction="Help the user with their order.",
    tools=[check_order_status, issue_refund, example_tool],
)
```

--------------------------------

### Install ElevenLabs SDK

Source: https://github.com/google/adk-python/blob/main/contributing/samples/live/cascade/README_elevenlabs.md

Install the google-adk package with the elevenlabs extra to get a supported elevenlabs release and websockets for the realtime recognizer.

```bash
pip install "google-adk[elevenlabs]"
```

--------------------------------

### Set up environment and install dependencies

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/authn-adk-all-in-one/README.md

Navigate to the integration sample directory, create a Python virtual environment, activate it, and install required packages.

```bash
# Go to the cloned directory
cd adk-python
# Navigate to the all in one authentication sample
cd contributing/samples/integrations/authn-adk-all-in-one/

python3 -m venv .venv

. .venv/bin/activate

pip install -r requirements.txt

```

--------------------------------

### Install dependencies for two-process worker setup

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/livekit/README.md

Installs the required packages for the two-process architecture, including google-adk with LiveKit support, livekit-agents, FastAPI, and Uvicorn.

```bash
pip install "google-adk[livekit]" "livekit-agents>=1.4" fastapi uvicorn
```

--------------------------------

### Clone and install Tau-bench repository

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/gepa/README.md

Clones the Tau-bench repository and installs it in editable mode for running Tau-bench integration examples.

```bash
git clone https://github.com/sierra-research/tau-bench.git
cd tau-bench/
pip install -e .
```

--------------------------------

### Install addlicense for license header checks

Source: https://github.com/google/adk-python/blob/main/CONTRIBUTING.md

Optionally installs Google's addlicense tool (requires Go). If not installed, the pre-commit hook is skipped and CI catches missing headers.

```shell
go install github.com/google/addlicense@latest
```

--------------------------------

### Start MCP server with client verification via filesystem_server.py

Source: https://github.com/google/adk-python/blob/main/contributing/samples/mcp/mcp_sse_mtls_agent/README.md

Sets environment variables to point to the certificate configuration and test CA, then starts the server on https://localhost:3000.

```bash
# Point to the certificate config
export GOOGLE_API_CERTIFICATE_CONFIG=$(pwd)/certificate_config.json

# Tell the server to trust our test CA for client verification
export SSL_CA_CERTS=$(pwd)/ca.crt

# Run the server
python filesystem_server.py
```

--------------------------------

### Install google-adk with Slack support

Source: https://github.com/google/adk-python/blob/main/src/google/adk/integrations/slack/README.md

Installs the google-adk library with the required slack dependencies.

```bash
pip install "google-adk[slack]"
```

--------------------------------

### Start Toolbox Server with --tools-file

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/toolbox_agent/README.md

Starts the Toolbox server on `http://127.0.0.1:5000` using the specified YAML configuration file.

```bash
toolbox --tools-file "tools.yaml"
```

--------------------------------

### Implement BaseExampleProvider for per-query example selection

Source: https://github.com/google/adk-python/blob/main/docs/guides/examples/example/index.md

Defines a custom provider that selects examples by intent from an in-memory dictionary. The get_examples method is synchronous and receives raw user text, so avoid network calls.

```python
class IntentExampleProvider(BaseExampleProvider):

  def __init__(self, examples_by_intent: dict[str, list[Example]]):
    self._examples_by_intent = examples_by_intent
```

--------------------------------

### Quick start OAuth2 server and run weather agent

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/oauth2_client_credentials/README.md

Sequential steps to start the mock OAuth2 server in the background, configure environment variables in `.env`, and query the weather assistant.

```bash
python contributing/samples/integrations/oauth2_client_credentials/oauth2_test_server.py &
```

```bash
# Choose Model Backend: 0 -> ML Dev, 1 -> Vertex
GOOGLE_GENAI_USE_ENTERPRISE=1

# ML Dev backend config
GOOGLE_API_KEY=your_google_api_key_here

# Vertex backend config
GOOGLE_CLOUD_PROJECT=your_project_id
GOOGLE_CLOUD_LOCATION=us-central1
```

```bash
# Ask for weather
python contributing/samples/integrations/oauth2_client_credentials/main.py "What's the weather in Tokyo?"
```

--------------------------------

### Install dependencies for custom web client

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/gcp_auth/README.md

Navigates to the 3LO client directory and installs all required Python packages from requirements.txt.

```bash
cd contributing/samples/integrations/gcp_auth/client
pip install -r requirements.txt
```

--------------------------------

### Install dependencies with uv

Source: https://github.com/google/adk-python/blob/main/contributing/samples/models/grok/README.md

Installs the required OpenAI and ADK extension dependencies from the repository root.

```bash
uv sync --extra extensions
```

--------------------------------

### Install OpenAI extra dependencies

Source: https://github.com/google/adk-python/blob/main/contributing/samples/models/hello_world_openai/README.md

Run from the repository root to install the OpenAI extra required for the sample.

```bash
uv sync --extra extensions
```

--------------------------------

### Install LiveKit integration package

Source: https://github.com/google/adk-python/blob/main/src/google/adk/integrations/livekit/README.md

Install the Google ADK package with the LiveKit integration extra.

```bash
pip install "google-adk[livekit]"
```

--------------------------------

### Start the local MCP filesystem server with uv

Source: https://github.com/google/adk-python/blob/main/contributing/samples/mcp/mcp_streamablehttp_agent/README.md

Run this command to start the local MCP server before running the agent.

```bash
uv run filesystem_server.py
```

--------------------------------

### Install Dependencies and Configure Authentication

Source: https://github.com/google/adk-python/blob/main/contributing/samples/models/interactions_api/README.md

Sets up the virtual environment and authentication credentials via either Google Cloud Project or Google API Key.

```bash
# From the adk-python root directory
uv sync --all-extras
source .venv/bin/activate

# Set up authentication (choose one):
# Option 1: Using Google Cloud credentials
export GOOGLE_CLOUD_PROJECT=your-project-id

# Option 2: Using API Key
export GOOGLE_API_KEY=your-api-key
```

--------------------------------

### Install dependencies with uv

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/files_retrieval_agent/README.md

Installs the required project dependencies including all extras required for the files retrieval agent sample.

```bash
uv sync --all-extras
```

--------------------------------

### Start the hotel booking backend API

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/authn-adk-all-in-one/README.md

Activate the virtual environment in a separate shell and start the hotel booker application backend service.

```bash
# Go to the cloned directory
cd adk-python
# Navigate to the all in one authentication sample
cd contributing/samples/integrations/authn-adk-all-in-one/

# Activate Env for this shell
. .venv/bin/activate

cd hotel_booker_app/

# Start the hotel booker application
python main.py

```

--------------------------------

### Create virtual environment and install dependencies

Source: https://github.com/google/adk-python/blob/main/CONTRIBUTING.md

Creates a Python 3.11 virtual environment, activates it, and installs all dependencies including extras.

```shell
uv venv --python "python3.11" ".venv"
source .venv/bin/activate
uv sync --all-extras
```

--------------------------------

### Install google-adk with optional extensions

Source: https://github.com/google/adk-python/blob/main/README.md

Installs the google-adk package along with its optional extensions extra.

```bash
pip install "google-adk[extensions]"
```

--------------------------------

### Start ADK API Server for Live Sample

Source: https://github.com/google/adk-python/blob/main/contributing/samples/live/api_server_example/readme.md

Run this first for the folder that directly contains your agent folder. APP_NAME defaults to single_agent, which lives in contributing/samples/live. Keep this running.

```shell
adk api_server contributing/samples/live
```

--------------------------------

### Install LiteLLM

Source: https://github.com/google/adk-python/blob/main/contributing/samples/models/hello_world_nvidia/README.md

Install the LiteLLM dependency to enable ADK integration with NVIDIA NIM models.

```bash
pip install litellm
```

--------------------------------

### Run the live workflow with adk web

Source: https://github.com/google/adk-python/blob/main/contributing/samples/live/workflow/README.md

Starts the ADK web interface for the workflow sample directory so you can begin a Live Session.

```bash
uv run adk web contributing/samples/live/workflow
```

--------------------------------

### Start the ADK agent web interface

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/authn-adk-all-in-one/README.md

Activate the virtual environment, copy environment configuration, ensure the GOOGLE_API_KEY is configured, and launch the ADK agent web server.

```bash
# Go to the cloned directory
cd adk-python
# Navigate to the all in one authentication sample
cd contributing/samples/integrations/authn-adk-all-in-one/

# Activate Env for this shell
. .venv/bin/activate

cd adk_agents/

cp sample.env .env

# ⚠️ Make sure to update the API KEY (GOOGLE_API_KEY) in .env file

# Run the agent
adk web

```

--------------------------------

### Example PostgreSQL connection string formats

Source: https://github.com/google/adk-python/blob/main/contributing/samples/mcp/mcp_postgres_agent/README.md

Shows sample URI formats for connecting to a local PostgreSQL instance or a hosted database service such as Supabase.

```text
postgresql://username:password@localhost:5432/mydb
postgresql://postgres.xyz:password@aws-region.pooler.supabase.com:5432/postgres
```

--------------------------------

### Launch web UI for basic sample with adk web

Source: https://github.com/google/adk-python/blob/main/contributing/samples/plugins/plugin_reflect_tool_retry/README.md

Starts the ADK web interface for running and testing the basic retry plugin sample agent.

```bash
$ adk web contributing/samples/plugins/plugin_reflect_tool_retry
```

--------------------------------

### Start the Local MCP Header Server with uv

Source: https://github.com/google/adk-python/blob/main/contributing/samples/mcp/mcp_dynamic_header_agent/README.md

Starts the local MCP server before running the dynamic header agent.

```bash
uv run header_server.py
```

--------------------------------

### Running an agent with Runner and session setup (Before)

Source: https://github.com/google/adk-python/blob/main/contributing/samples/core/runner_debug_example/README.md

Standard multi-line execution pattern using InMemorySessionService and Runner.run_async with explicit session creation.

```python
from google.adk.sessions import InMemorySessionService
from google.genai import types

APP_NAME = "default"
USER_ID = "default"
session_service = InMemorySessionService()
runner = Runner(agent=agent, app_name=APP_NAME, session_service=session_service)
session = await session_service.create_session(
    app_name=APP_NAME, user_id=USER_ID, session_id="default"
)
content = types.Content(role="user", parts=[types.Part.from_text(text="Hi")])
async for event in runner.run_async(
    user_id=USER_ID, session_id=session.id, new_message=content
):
    if event.content and event.content.parts:
        print(event.content.parts[0].text)
```

--------------------------------

### Launch the ADK Web UI for Integration Samples

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/api_registry_agent/README.md

Starts the ADK web server to run and interact with the agent from a browser. Access the UI at http://127.0.0.1:8000 and choose the api_registry_agent.

```bash
adk web contributing/samples/integrations

```

--------------------------------

### Define sequential edges with START

Source: https://github.com/google/adk-python/blob/main/docs/guides/workflow/workflow/index.md

Use a tuple to run steps sequentially after START.

```python
edges=[(START, step_a, step_b)]
```

--------------------------------

### Install the locally built wheel file

Source: https://github.com/google/adk-python/blob/main/CONTRIBUTING.md

Installs the wheel file built by uv build into the active virtual environment. Replace <version> with the actual version.

```shell
pip install dist/google_adk-<version>-py3-none-any.whl
```

--------------------------------

### Install google-adk with e2b extra

Source: https://github.com/google/adk-python/blob/main/contributing/samples/environment_and_skills/e2b_environment/README.md

Install the Google ADK package along with the E2B integration dependencies.

```bash
pip install google-adk[e2b]
```

--------------------------------

### Install dependencies with pip

Source: https://github.com/google/adk-python/blob/main/contributing/samples/mcp/mcp_server_side_sampling/README.md

Install the required Python packages for running FastMCP server-side sampling.

```bash
pip install fastmcp openai litellm
```

--------------------------------

### Start the ADK Web Server with adk web

Source: https://github.com/google/adk-python/blob/main/contributing/samples/live/single_agent/readme.md

Run from the root directory that contains the single_agent folder. After starting, open the URL shown in the terminal (typically http://localhost:8000) to access the ADK Web UI.

```bash
adk web
```

--------------------------------

### Install google-adk with Daytona extra

Source: https://github.com/google/adk-python/blob/main/contributing/samples/environment_and_skills/daytona_environment/README.md

Installs the google-adk package with the optional Daytona integration dependencies.

```bash
pip install google-adk[daytona]
```

--------------------------------

### Execute Runner Debug Example Module

Source: https://github.com/google/adk-python/blob/main/contributing/samples/core/runner_debug_example/README.md

Runs the runner debug example main script containing eight demonstration scenarios.

```bash
python -m contributing.samples.core.runner_debug_example.main
```

--------------------------------

### Launch workflow samples in Web UI

Source: https://github.com/google/adk-python/blob/main/contributing/samples/workflows/auth_oauth/README.md

Starts the ADK web interface to interactively test and run workflow samples.

```bash
adk web contributing/samples/workflows/
```

--------------------------------

### Install required packages for PostgreSQL DatabaseSessionService

Source: https://github.com/google/adk-python/blob/main/contributing/samples/context_management/postgres_session_service/README.md

Installs the required Python packages including google-adk, asyncpg async driver, and greenlet.

```bash
pip install google-adk asyncpg greenlet

```

--------------------------------

### Launch the ADK web interface

Source: https://github.com/google/adk-python/blob/main/contributing/samples/mcp/mcp_in_agent_tool_remote/README.md

Starts the ADK web interface pointing to the MCP sample directory. Select mcp_in_agent_tool_remote from the interface to interact with the agent.

```bash
adk web contributing/samples/mcp
```

--------------------------------

### Install dependencies with OpenAI Responses extra

Source: https://github.com/google/adk-python/blob/main/contributing/samples/models/azure_responses_streaming/README.md

Install required dependencies from the repository root, including the OpenAI Responses extensions extra.

```bash
uv sync --extra extensions
```

--------------------------------

### Start agent in interactive mode with adk web

Source: https://github.com/google/adk-python/blob/main/contributing/samples/adk_team/adk_answering_agent/README.md

Starts a local server and renders the agent's interactive web interface in a browser. Ensure all required environment variables are set before executing.

```bash
adk web
```

--------------------------------

### Install ADK with Slack Support via pip

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/slack_agent/README.md

Install the ADK package with the Slack integration dependencies included.

```bash
pip install "google-adk[slack]"

```

--------------------------------

### Complete SKILL.md file example

Source: https://github.com/google/adk-python/blob/main/docs/guides/skills/skill/index.md

Defines a complete SKILL.md file containing YAML frontmatter configuration and model instruction steps that reference local skill files and tools.

```markdown
---
name: weather-skill
description: A skill that provides weather information based on reference data and scripts.
metadata:
  adk_additional_tools:
    - get_wind_speed
---

Step 1: Check 'references/weather_info.md' for the current weather.
Step 2: If humidity is requested, run 'scripts/get_humidity.py' with the `location` argument.
Step 3: If wind speed is requested, use the `get_wind_speed` tool.
Step 4: Provide the complete weather update to the user.
```

--------------------------------

### Install Python dependencies with uv pip

Source: https://github.com/google/adk-python/blob/main/contributing/samples/multimodal/computer_use/README.md

Installs the required Python packages for the computer use agent from the sample requirements file.

```bash
uv pip install -r contributing/samples/computer_use/requirements.txt

```

--------------------------------

### Install pre-commit and tox with tox-uv

Source: https://github.com/google/adk-python/blob/main/CONTRIBUTING.md

Installs pre-commit for code formatting and license enforcement, and tox with tox-uv for isolated multi-version testing.

```shell
uv tool install pre-commit
uv tool install tox --with tox-uv
```

--------------------------------

### Launch sample in interactive mode with adk run

Source: https://github.com/google/adk-python/blob/main/contributing/samples/multimodal/static_non_text_content/README.md

Starts the sample using ADK's built-in interactive CLI mode for free-form conversations.

```bash
cd contributing/samples
adk run static_non_text_content
```

--------------------------------

### Start PostgreSQL with Docker Compose

Source: https://github.com/google/adk-python/blob/main/contributing/samples/context_management/postgres_session_service/README.md

Starts a local PostgreSQL instance in detached mode using Docker Compose.

```bash
docker compose up -d
```

--------------------------------

### Use VertexAiExampleStore with ExampleTool

Source: https://github.com/google/adk-python/blob/main/docs/guides/examples/example/index.md

Creates an ExampleTool backed by a Vertex AI Example Store. Each turn runs a similarity search with hard-coded top_k=10 and score floor 0.5. Requires the store resource name.

```python
example_tool = ExampleTool(
    VertexAiExampleStore(
        "projects/my-project/locations/us-central1/exampleStores/my-store"
    )
)
```

--------------------------------

### Install pymongo separately

Source: https://github.com/google/adk-python/blob/main/src/google/adk/integrations/mongodb/README.md

Installs pymongo if ADK is already installed.

```bash
pip install "pymongo>=4.9,<5"
```

--------------------------------

### Install Tau-bench and GEPA dependencies

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/gepa/gepa_tau_bench.ipynb

Clones the Google ADK and Tau-bench repositories, installs them in editable mode, and installs `gepa` and `retry` packages.

```python
# @title Install Tau-bench and GEPA
!git clone https://github.com/google/adk-python.git
!git clone https://github.com/sierra-research/tau-bench.git
!pip install -e ./adk-python --quiet

%cd tau-bench/
!pip install -e . --quiet

%cd ..
!pip install gepa --quiet

!pip install retry --quiet
```

--------------------------------

### Launch the ADK Dev UI

Source: https://github.com/google/adk-python/blob/main/contributing/samples/models/grok/README.md

Starts the local ADK web interface to interactively test the Grok agent.

```bash
uv run --extra extensions adk web contributing/samples/models/grok
```

--------------------------------

### Start remote greet_agent A2A server with adk api_server --a2a

Source: https://github.com/google/adk-python/blob/main/contributing/samples/a2a/a2a_state_forwarding/README.md

Run this in one terminal to start the remote greet_agent as an A2A server on port 8001. The main agent is run separately afterward.

```bash
adk api_server --a2a --port 8001 contributing/samples/a2a/a2a_state_forwarding/remote_a2a
```

--------------------------------

### Pass plain dictionaries to ExampleTool for text-only examples

Source: https://github.com/google/adk-python/blob/main/docs/guides/examples/example/index.md

ExampleTool accepts plain dictionaries and validates them into Example objects, which is shorter when the examples are all text.

```python
example_tool = ExampleTool([
    {
        "input": {"role": "user", "parts": [{"text": "Is 7 a prime number?"}]},
        "output": [{"role": "model", "parts": [{"text": "Yes, 7 is a prime number."}]}],
    },
])
```

--------------------------------

### Create a fan-out workflow with START, JoinNode, and report

Source: https://github.com/google/adk-python/blob/main/docs/guides/workflow/graph/index.md

Complete workflow where START fans out to word_count and shout running in parallel, JoinNode waits for both, and report reads the aggregated result. The single chain tuple expands into five edges, and both workers receive the same user message because they follow START.

```python
from typing import Any

from google.adk import Workflow
from google.adk.workflow import JoinNode, START

def word_count(node_input: str) -> int:
    return len(node_input.split())

def shout(node_input: str) -> str:
    return node_input.upper()

join = JoinNode(name="join")

def report(node_input: dict[str, Any]) -> str:
    return f"{node_input['shout']} ({node_input['word_count']} words)"

workflow = Workflow(
    name="fan_out_workflow",
    edges=[
        (START, (word_count, shout), join, report),
    ],
)
```

--------------------------------

### Create a voice assistant agent with CascadeLive, GeminiSTT, and GeminiTTS

Source: https://github.com/google/adk-python/blob/main/docs/guides/live/gemini_tts/index.md

Pass GeminiTTS as the tts of CascadeLive with the voice name in speech_config; this example uses the Puck voice and GeminiSTT for input. The instruction requests short sentences ending in punctuation because GeminiTTS starts speaking at the first sentence end. Audio reaches the client as 16-bit mono PCM blobs labeled audio/pcm;rate=24000.

```python
from google.adk import Agent
from google.adk.live import CascadeLive
from google.adk.live import GeminiSTT
from google.adk.live import GeminiTTS

agent = Agent(
    name='voice_assistant',
    model=CascadeLive(
        model='gemini-3.8-flash',
        stt=GeminiSTT(),
        tts=GeminiTTS(speech_config='Puck'),
    ),
    instruction=(
        'You are a helpful voice assistant. Answer in one or two short'
        ' sentences, and end every sentence with punctuation.'
    ),
)
```

--------------------------------

### Launch Agent Builder Assistant via ADK Web Interface

Source: https://github.com/google/adk-python/blob/main/src/google/adk/cli/built_in_agents/README.md

Run this command from the ADK project root to start the web interface for the Agent Builder Assistant.

```bash
# From the ADK project root
adk web src/google/adk/agent_builder_assistant
```

--------------------------------

### Start the local OAuth2 mock test server

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/oauth2_client_credentials/README.md

Runs the mock OAuth2 server to test client credentials token exchange and provide protected weather endpoints.

```bash
python contributing/samples/integrations/oauth2_client_credentials/oauth2_test_server.py
```

--------------------------------

### Install ADK with MongoDB extra

Source: https://github.com/google/adk-python/blob/main/src/google/adk/integrations/mongodb/README.md

Installs ADK with the mongodb extra, which pulls in pymongo.

```bash
pip install google-adk[mongodb]
```

--------------------------------

### Start MCP simple-tool server in SSE mode

Source: https://github.com/google/adk-python/blob/main/contributing/samples/mcp/mcp_in_agent_tool_remote/README.md

Starts the remote MCP simple-tool server using uvx on port 3000 using the SSE transport. Run this in a separate terminal before launching the ADK interface.

```bash
# Run the server using uvx (no installation needed)
# Port 3000 avoids conflict with adk web (which uses 8000)
uvx --from 'git+https://github.com/modelcontextprotocol/python-sdk.git#subdirectory=examples/servers/simple-tool' \
    mcp-simple-tool --transport sse --port 3000
```

--------------------------------

### Install google-adk with GCP extras

Source: https://github.com/google/adk-python/blob/main/docs/guides/integrations/model_armor/index.md

Install the dependency required for Model Armor integration.

```shell
pip install 'google-adk[gcp]'
```

--------------------------------

### Download and Install MCP Toolbox Binary

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/toolbox_agent/README.md

Downloads the MCP Toolbox binary for the specified target OS and makes the binary executable.

```bash
export OS="linux/amd64" # one of linux/amd64, darwin/arm64, darwin/amd64, or windows/amd64
curl -O https://storage.googleapis.com/mcp-toolbox-for-databases/v1.1.0/$OS/toolbox
chmod +x toolbox
```

--------------------------------

### Install OpenSandbox extra for google-adk

Source: https://github.com/google/adk-python/blob/main/contributing/samples/environment_and_skills/opensandbox_environment/README.md

Installs the OpenSandbox extra required before using OpenSandboxEnvironment.

```bash
pip install google-adk[opensandbox]
```

--------------------------------

### Using run_debug with various options and configurations

Source: https://github.com/google/adk-python/blob/main/contributing/samples/core/runner_debug_example/README.md

Examples demonstrating minimal usage, multiple queries, custom session credentials, quiet event capture, verbose tool logging, and custom RunConfig.

```python
# Minimal usage
runner = InMemoryRunner(agent=agent)
await runner.run_debug("What's the weather?")

# Multiple queries
await runner.run_debug(["Query 1", "Query 2", "Query 3"])

# Custom session
await runner.run_debug(
    "Hello",
    user_id="alice",
    session_id="debug_session"
)

# Capture events without printing
events = await runner.run_debug(
    "Process this",
    quiet=True
)

# Show tool calls with verbose mode
await runner.run_debug(
    "What's the weather?",
    verbose=True  # Shows [Calling tool: ...] and [Tool result: ...]
)

# With custom configuration
from google.adk.agents.run_config import RunConfig
config = RunConfig(max_llm_calls=10)
await runner.run_debug("Query", run_config=config)
```

--------------------------------

### Run an ADK Agent with SlackRunner

Source: https://github.com/google/adk-python/blob/main/src/google/adk/integrations/slack/README.md

Initializes Slack AsyncApp and SlackRunner, then starts listening for events in Socket Mode using SLACK_BOT_TOKEN and SLACK_APP_TOKEN.

```python
import asyncio
import os
from google.adk.runners import Runner
from google.adk.integrations.slack import SlackRunner
from slack_bolt.app.async_app import AsyncApp

async def main():
    # 1. Initialize your ADK Runner (with your agent)
    # runner = Runner(agent=my_agent, session_service=my_session_service)

    # 2. Initialize Slack AsyncApp with your Bot Token
    slack_app = AsyncApp(token=os.environ["SLACK_BOT_TOKEN"])

    # 3. Initialize the SlackRunner
    slack_runner = SlackRunner(runner=runner, slack_app=slack_app)

    # 4. Start the runner in Socket Mode with your App Token
    await slack_runner.start(app_token=os.environ["SLACK_APP_TOKEN"])

if __name__ == "__main__":
    asyncio.run(main())
```

--------------------------------

### Install Agent Identity dependencies

Source: https://github.com/google/adk-python/blob/main/src/google/adk/integrations/agent_identity/README.md

Installs the Google ADK package with the agent-identity integration extras.

```bash
pip install "google-adk[agent-identity]"
```

--------------------------------

### Start Remote A2A Agent Server with Uvicorn

Source: https://github.com/google/adk-python/blob/main/contributing/samples/a2a/a2a_root/README.md

Starts the remote Hello World A2A agent service on localhost port 8001 using uvicorn before running the main agent.

```bash
# Start the remote agent using uvicorn
uvicorn contributing.samples.a2a.a2a_root.remote_a2a.hello_world.agent:a2a_app --host localhost --port 8001
```

--------------------------------

### Install ADK and GEPA dependencies with pip

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/gepa/README.md

Installs the Google ADK along with required dependencies including GEPA, Jinja2, LiteLLM, Abseil, and retry.

```bash
pip install google-adk absl-py gepa jinja2 litellm retry
```

--------------------------------

### Configure and start the IDP service

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/authn-adk-all-in-one/README.md

Generate an RSA key pair, extract the public key to configure JWKS and private key in the environment, and run the Identity Provider application.

```bash
cd idp

# Create .env file by copying the existing one.
cp sample.env .env
cp sample.jwks.json jwks.json


# Carry out following steps
# 1. Generate a key pair, When asked about passphrase please press enter (empty passphrase)
ssh-keygen -t rsa -b 2048 -m PEM -f private_key.pem

# 2. Extract the public key
openssl rsa -in private_key.pem -pubout > pubkey.pub

# 3. Generate the jwks.json content using https://jwkset.com/generate and this public key (choose key algorithm RS256 and Key use Signature) (Please check the screenshot)
# 4. Update the jwks.json with the key jwks key created in 3 (please check the screenshot)
# 5. Update the env file with the private key
cat private_key.pem | tr -d "\n"
# 6. Carefully copy output of the command above into the .env file to update the value of PRIVATE_KEY
# 7. save jwks.json and .env

# Start the IDP
python app.py
```

--------------------------------

### Example query prompt for browser task

Source: https://github.com/google/adk-python/blob/main/contributing/samples/multimodal/computer_use/README.md

Sample prompt instructing the agent to navigate to Google Flights and search for flights.

```text
find me a flight from SF to Hawaii on next Monday, coming back on next Friday. start by navigating directly to flights.google.com

```

--------------------------------

### Install Chromium browser for Playwright

Source: https://github.com/google/adk-python/blob/main/contributing/samples/multimodal/computer_use/README.md

Downloads and installs the Chromium browser binary used by Playwright.

```bash
playwright install chromium

```

--------------------------------

### Pinned setuptools, simple-websocket, six, and slack packages

Source: https://github.com/google/adk-python/blob/main/constraints-3.15.txt

setuptools is required by llama-index-core, simple-websocket by python-engineio, six by kubernetes and others, and slack-sdk by slack-bolt.

```text
setuptools==84.0.0
    # via
    #   -c constraints-3.15.txt.stable.tmp
    #   llama-index-core
simple-websocket==1.1.0
    # via
    #   -c constraints-3.15.txt.stable.tmp
    #   python-engineio
six==1.17.0
    # via
    #   -c constraints-3.15.txt.stable.tmp
    #   kubernetes
    #   python-dateutil
    #   rouge-score
slack-bolt==1.30.0
    # via
    #   -c constraints-3.15.txt.stable.tmp
    #   google-adk (pyproject.toml)
slack-sdk==3.45.0
    # via
    #   -c constraints-3.15.txt.stable.tmp
    #   slack-bolt
```

--------------------------------

### Run Live Agent Example Client

Source: https://github.com/google/adk-python/blob/main/contributing/samples/live/api_server_example/readme.md

Run this in a separate window after the API server is running. Ensure the terminal has mic/sound device permission; native terminals may work better than VSCode terminals.

```shell
python3 agent_example.py
```

--------------------------------

### Install google-adk with constraints file for Python 3.11

Source: https://github.com/google/adk-python/blob/main/README.md

Downloads the constraints file for Python 3.11, installs google-adk with it, then removes the file. Use the constraints file matching your Python version (3.11 to 3.15).

```bash
# For example, for Python 3.11
curl -o constraints-3.11.txt https://raw.githubusercontent.com/google/adk-python/main/constraints-3.11.txt
pip install google-adk -c constraints-3.11.txt
rm constraints-3.11.txt
```

--------------------------------

### Create a basic Agent with google.adk.Agent

Source: https://github.com/google/adk-python/blob/main/README.md

Defines a root agent with a name, model, and instruction. Use this as the starting point for a simple conversational agent.

```python
from google.adk import Agent

root_agent = Agent(
    name="greeting_agent",
    model="gemini-3.5-flash",
    instruction="You are a helpful assistant. Greet the user warmly.",
)
```

--------------------------------

### Example interaction: YouTube search without authentication

Source: https://github.com/google/adk-python/blob/main/contributing/samples/a2a/a2a_auth/README.md

Shows sample root agent interaction delegating to the YouTube Search Agent without requiring user authentication.

```text
User: Search for 3 Taylor Swift music videos
Agent: I'll help you search for Taylor Swift music videos on YouTube.
[Agent delegates to YouTube Search Agent]
Agent: I found 3 Taylor Swift music videos:
1. "Anti-Hero" - Official Music Video
2. "Shake It Off" - Official Music Video
3. "Blank Space" - Official Music Video
```

--------------------------------

### Set Google API Key Environment Variable

Source: https://github.com/google/adk-python/blob/main/contributing/samples/core/runner_debug_example/README.md

Sets the required GOOGLE_API_KEY environment variable before executing the example.

```bash
export GOOGLE_API_KEY="your-api-key"
```

--------------------------------

### Install CrewAI Tools Dependency

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/crewai_tool_kwargs/README.md

Installs the required crewai-tools package (requires Python 3.10+).

```bash
pip install 'crewai-tools>=0.2.0'
```

--------------------------------

### Run the basic plugin sample agent

Source: https://github.com/google/adk-python/blob/main/contributing/samples/plugins/plugin_basic/README.md

Execute the main.py script to run the sample agent with the basic plugin configured. Output displays plugin lifecycle counter logs alongside agent events.

```bash
python3 -m contributing.samples.plugins.plugin_basic.main
```

```bash
[Plugin] Agent run count: 1
[Plugin] LLM request count: 1
** Got event from hello_world
Hello world: query is [hello world]
** Got event from hello_world
[Plugin] LLM request count: 2
** Got event from hello_world
```

--------------------------------

### Start local LiveKit dev server

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/livekit/README.md

Runs a local LiveKit server in development mode with fixed public credentials in a dedicated terminal.

```bash
livekit-server --dev
```

--------------------------------

### Launch ADK Web UI for MCP Toolset Auth Sample

Source: https://github.com/google/adk-python/blob/main/contributing/samples/mcp/mcp_toolset_auth/README.md

Run the ADK web interface to test the sample interactively and provide authentication credentials manually when prompted.

```bash
adk web contributing/samples/mcp/mcp_toolset_auth
```

--------------------------------

### Install MongoDB extra for google-adk

Source: https://github.com/google/adk-python/blob/main/docs/guides/integrations/mongodb/mongodb_toolset/index.md

Installs the MongoDB integration dependencies, including pymongo, via pip.

```shell
pip install 'google-adk[mongodb]'
```

--------------------------------

### Start the LiveKit token and dispatch server

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/livekit/README.md

Runs the token and dispatch server in a separate terminal to handle dispatches and serve the web client interface at http://localhost:8080.

```bash
python -m contributing.samples.integrations.livekit.client.main
```

--------------------------------

### Build the wheel file

Source: https://github.com/google/adk-python/blob/main/CONTRIBUTING.md

Builds the distribution wheel file using uv.

```shell
uv build
```

--------------------------------

### Install pre-commit git hooks

Source: https://github.com/google/adk-python/blob/main/CONTRIBUTING.md

Installs the git hooks to automatically format and check code before committing.

```shell
pre-commit install
```

--------------------------------

### Set OLLAMA_API_BASE and launch ADK web interface

Source: https://github.com/google/adk-python/blob/main/contributing/samples/models/hello_world_ollama/README.md

Sets the OLLAMA_API_BASE environment variable to point to the local Ollama server and starts the ADK web interface.

```bash
export OLLAMA_API_BASE="http://localhost:11434"
adk web
```

--------------------------------

### Install stable google-adk with pip

Source: https://github.com/google/adk-python/blob/main/README.md

Installs the latest stable release of ADK. Requires Python 3.11+.

```bash
pip install google-adk
```

--------------------------------

### Configure Agent with CascadeLive, ElevenLabsSTT, and ElevenLabsTTS

Source: https://github.com/google/adk-python/blob/main/docs/guides/live/cascade_live/index.md

Sets up a voice assistant using gemini-3.5-flash as the text reasoner, with ElevenLabsSTT for ingress and ElevenLabsTTS for egress. The agent handles live voice input, executes tools during reasoning, and speaks the response back.

```python
agent = Agent(
    name='weather_assistant',
    model=CascadeLive(
        model='gemini-3.5-flash',
        stt=ElevenLabsSTT(),
        tts=ElevenLabsTTS(),
    ),
    instruction='You are a helpful voice assistant. Answer briefly.',
    tools=[get_current_weather],
)
```

--------------------------------

### Install Playwright system dependencies for Chromium

Source: https://github.com/google/adk-python/blob/main/contributing/samples/multimodal/computer_use/README.md

Installs the necessary system-level dependencies required by Chromium for Playwright automation.

```bash
playwright install-deps chromium

```

--------------------------------

### Launch MCP simple-tool server using uvx

Source: https://github.com/google/adk-python/blob/main/contributing/samples/mcp/mcp_in_agent_tool_stdio/README.md

Runs the MCP simple-tool server directly from the GitHub repository's subdirectory using uvx over a stdio connection.

```bash
uvx --from 'git+https://github.com/modelcontextprotocol/python-sdk.git#subdirectory=examples/servers/simple-tool' \
    mcp-simple-tool
```

--------------------------------

### Install google-adk dependencies

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/gcp_auth/README.md

Install the google-adk package including optional extras for Agent Identity and MCP support.

```bash
pip install "google-adk[agent-identity,mcp]"
```

--------------------------------

### Install google-adk GCP extra dependency

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/eventarc/generic_agent/README.md

Installs the google-adk package with the gcp extra required for Eventarc publishing.

```bash
pip install "google-adk[gcp]"
```

--------------------------------

### Install GCP Extra Dependency for ADK

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/eventarc/domain_specific_agent/README.md

Installs the google-adk package with GCP extras required for Eventarc publishing.

```bash
pip install "google-adk[gcp]"

```

--------------------------------

### Activate virtual environment

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/gcp_auth/README.md

Navigate to the project directory, initialize a Python virtual environment, and activate it.

```bash
cd adk-python
python3 -m venv .venv
source .venv/bin/activate
```

--------------------------------

### Install agent dependencies with pip

Source: https://github.com/google/adk-python/blob/main/contributing/samples/adk_team/adk_documentation/adk_release_analyzer/README.md

Upgrades pip and installs the required google-adk package along with database dependencies.

```bash
pip install --upgrade pip
pip install google-adk[db]
```

--------------------------------

### Example interaction: simple dice rolling

Source: https://github.com/google/adk-python/blob/main/contributing/samples/a2a/a2a_basic/README.md

Shows an example dialogue interacting with the root agent to roll a die.

```text
User: Roll a 6-sided die
Bot: I rolled a 4 for you.
```

--------------------------------

### Initializing Runner with an App or bare agent

Source: https://github.com/google/adk-python/blob/main/docs/guides/apps/app/index.md

Demonstrates current initialization using an App instance versus legacy agent wrapping. Passing an App preserves application-wide configuration and applies validation that bare agent initialization skips.

```python
# Current: the App carries the application-wide configuration.
runner = Runner(app=app, session_service=session_service)

# Legacy: the agent is wrapped in an App for you.
runner = Runner(
    app_name="weather_app",
    agent=root_agent,
    session_service=session_service,
)
```

--------------------------------

### Install test-specific packages with pip

Source: https://github.com/google/adk-python/blob/main/tests/integration/integrations/agent_identity/README.md

Install the google-cloud-iamconnectorcredentials package needed specifically for GCP Agent Identity integration tests.

```bash
pip install google-cloud-iamconnectorcredentials
```

--------------------------------

### Run the Files Retrieval Agent

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/files_retrieval_agent/README.md

Changes into the integrations sample directory and launches the agent via either the interactive CLI or the Web UI.

```bash
cd contributing/samples/integrations

# Interactive CLI
adk run files_retrieval_agent

# Web UI
adk web .
```

--------------------------------

### Install ADK with OpenAI extra via pip

Source: https://github.com/google/adk-python/blob/main/src/google/adk/integrations/openai/README.md

Installs the google-adk package along with required OpenAI integration dependencies.

```bash
pip install "google-adk[openai]"
```

--------------------------------

### Install ADK Toolbox Dependency with pip

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/toolbox_agent/README.md

Installs the required `toolbox-adk` extension needed by ADK's `ToolboxToolset` class.

```bash
pip install google-adk[toolbox]
```

--------------------------------

### Sample File Structure

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/crewai_tool_kwargs/README.md

Directory layout showing key files in the crewai_tool_kwargs integration sample.

```text
crewai_tool_kwargs/
├── __init__.py       # Module initialization
├── agent.py          # Agent with CrewAI tool
├── main.py           # Happy path test
└── README.md         # This file
```

--------------------------------

### Install GEPA and Dependencies

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/gepa/voter_agent/gepa.ipynb

Installs the Google ADK Python repository, GEPA, litellm, and retry packages in a Colab environment.

```python
# @title Install GEPA
!git clone https://github.com/google/adk-python.git
!pip install -e ./adk-python --quiet
!pip install gepa --quiet
!pip install litellm --quiet
!pip install retry --quiet
```

--------------------------------

### Run the non-blocking tool agent sample with adk web

Source: https://github.com/google/adk-python/blob/main/contributing/samples/live/non_blocking_tool_agent/README.md

Launches the sample in the ADK web interface. Start a Live Session and trigger the slow_background_task tool to observe the model continuing to converse while the tool runs.

```bash
uv run adk web contributing/samples/live/non_blocking_tool_agent
```

--------------------------------

### Install agent dependencies with pip

Source: https://github.com/google/adk-python/blob/main/contributing/samples/adk_team/adk_answering_agent/README.md

Upgrades pip and installs core libraries required by the agent, including google-adk and google-cloud-discoveryengine.

```bash
pip install --upgrade pip
pip install google-adk google-cloud-discoveryengine
```

--------------------------------

### Initialize SQLite Database with sqlite3 CLI

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/toolbox_agent/README.md

Creates and opens a new SQLite database file named `tool_box.db`.

```bash
sqlite3 tool_box.db
```

--------------------------------

### Copy sample environment configuration file

Source: https://github.com/google/adk-python/blob/main/contributing/samples/models/hello_world_apigeellm/README.md

Copies the template environment file to create a local .env configuration file in the project directory.

```bash
cp .env-sample .env
```

--------------------------------

### Example interaction: combined operations

Source: https://github.com/google/adk-python/blob/main/contributing/samples/a2a/a2a_basic/README.md

Shows an example dialogue combining dice rolling with remote prime number verification.

```text
User: Roll a 10-sided die and check if it's prime
Bot: I rolled an 8 for you.
Bot: 8 is not a prime number.
```

--------------------------------

### Create and run an LlmAgent with a weather tool and dynamic instructions

Source: https://github.com/google/adk-python/blob/main/docs/guides/agents/llm_agent/index.md

Configures an LlmAgent with a weather tool and dynamic instruction placeholder, then executes it within an App and Runner. The runner resolves {user_name} from session state and executes tool calls before returning responses.

```python
import asyncio

from google.adk.agents import LlmAgent
from google.adk.apps import App
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService
from google.genai import types


def get_current_weather(city: str) -> str:
  """Returns current weather conditions for a given city.

  Args:
    city: The city name.
  """
  return f"The weather in {city} is sunny and 22°C."


async def main() -> None:
  # Configure LlmAgent with instructions, model, and tools
  agent = LlmAgent(
      name="weather_assistant",
      model="gemini-2.5-flash",
      instruction="You are a helpful assistant for user {user_name}.",
      tools=[get_current_weather],
  )

  app = App(name="weather_app", root_agent=agent)
  session_service = InMemorySessionService()
  runner = Runner(app=app, session_service=session_service)

  session = await session_service.create_session(
      app_name="weather_app",
      user_id="user_42",
      state={"user_name": "Alice"},
  )

  message = types.Content(
      role="user",
      parts=[types.Part.from_text(text="What is the weather in Paris?")],
  )

  async for event in runner.run_async(
      user_id="user_42",
      session_id=session.id,
      new_message=message,
  ):
    if event.content and event.content.parts:
      for part in event.content.parts:
        if part.text:
          print("Output:", part.text)


if __name__ == "__main__":
  asyncio.run(main())
```

--------------------------------

### Install youtube_search dependency

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/langchain_youtube_search_agent/README.md

Installs the required `youtube_search` dependency via `uv pip` to enable YouTube video search with Langchain.

```python
uv pip install youtube_search
```

--------------------------------

### Instantiate an Agent with tools using google.adk.agents.llm_agent.Agent

Source: https://github.com/google/adk-python/blob/main/contributing/samples/core/quickstart/README.md

Configure an agent instance by passing tool functions to the tools parameter list along with descriptive instructions.

```python
from google.adk.agents.llm_agent import Agent

root_agent = Agent(
    name="weather_time_agent",
    description=(
        "Agent to answer questions about the time and weather in a city."
    ),
    instruction=(
        "I can answer your questions about the time and weather in a city."
    ),
    tools=[get_weather, get_current_time],
)
```

--------------------------------

### Initialize and create a session with DatabaseSessionService

Source: https://github.com/google/adk-python/blob/main/docs/guides/sessions/session/index.md

Initialize DatabaseSessionService using an async database driver URL within an async context manager. As an async context manager, it handles closing the owned engine on exit. Calling prepare_tables() is optional and will otherwise occur on first use.

```python
from google.adk.sessions import DatabaseSessionService

async with DatabaseSessionService("sqlite+aiosqlite:///./sessions.db") as svc:
  await svc.prepare_tables()  # optional; otherwise done on first use
  session = await svc.create_session(app_name=APP_NAME, user_id=USER_ID)
```

--------------------------------

### Install upload script dependencies with pip

Source: https://github.com/google/adk-python/blob/main/contributing/samples/adk_team/adk_answering_agent/README.md

Installs additional Python packages required specifically to run the document upload script.

```bash
pip install google-cloud-storage markdown
```

--------------------------------

### Example NVIDIA NIM model identifiers

Source: https://github.com/google/adk-python/blob/main/contributing/samples/models/hello_world_nvidia/README.md

List example model identifiers supported by NVIDIA NIM, formatted with the nvidia_nim/ prefix.

```python
# Examples of available models
models = [
    "nvidia_nim/meta/llama3-8b-instruct",
    "nvidia_nim/meta/llama3-70b-instruct",
    "nvidia_nim/microsoft/phi-3-medium-4k-instruct",
    "nvidia_nim/mistralai/mistral-7b-instruct",
    "nvidia_nim/google/gemma-7b",
    "nvidia_nim/nvidia/nemotron-4-340b-instruct",
    # ... and many more
]
```

--------------------------------

### Example interaction: prime number checking

Source: https://github.com/google/adk-python/blob/main/contributing/samples/a2a/a2a_basic/README.md

Shows an example dialogue delegating a prime number check to the remote A2A agent.

```text
User: Is 7 a prime number?
Bot: Yes, 7 is a prime number.
```

--------------------------------

### Run interactive demo using ADK commands

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/oauth2_client_credentials/README.md

Launches the OAuth2 weather integration either through an interactive terminal session or a local web interface.

```bash
# Interactive CLI
adk run contributing/samples/integrations/oauth2_client_credentials

# Interactive web UI
adk web contributing/samples/integrations
```

--------------------------------

### Start the Remote Approval Agent server

Source: https://github.com/google/adk-python/blob/main/contributing/samples/a2a/a2a_human_in_loop/README.md

Start the remote A2A server serving the human-in-the-loop approval agent on port 8001.

```bash
# Start the remote a2a server that serves the human-in-the-loop approval agent on port 8001
adk api_server --a2a --port 8001 contributing/samples/a2a/a2a_human_in_loop/remote_a2a
```

--------------------------------

### Launch interactive web interface with adk web

Source: https://github.com/google/adk-python/blob/main/contributing/samples/adk_team/adk_documentation/adk_release_analyzer/README.md

Starts a local web server and interface for running the agent interactively to review recommendations before issue creation. Ensure the INTERACTIVE environment variable is set to 1 or left unset.

```bash
adk web contributing/samples/adk_team/adk_documentation
```

--------------------------------

### Run session state sample agent with replay

Source: https://github.com/google/adk-python/blob/main/contributing/samples/context_management/session_state_agent/README.md

Execute the session state persistence demo agent using the ADK CLI replay flag with input.json and observe the callback lifecycle assertions in the output.

```bash
$ adk run contributing/samples/context_management/session_state_agent --replay contributing/samples/context_management/session_state_agent/input.json
```

```bash
[user]: hello world!
===================== In before_agent_callback ==============================
** Asserting keys are cached in context: ['before_agent_callback_state_key'] pass ✅
** Asserting keys are already persisted in session: [] pass ✅
** Asserting keys are not persisted in session yet: ['before_agent_callback_state_key'] pass ✅
============================================================
===================== In before_model_callback ==============================
** Asserting keys are cached in context: ['before_agent_callback_state_key', 'before_model_callback_state_key'] pass ✅
** Asserting keys are already persisted in session: ['before_agent_callback_state_key'] pass ✅
** Asserting keys are not persisted in session yet: ['before_model_callback_state_key'] pass ✅
============================================================
===================== In after_model_callback ==============================
** Asserting keys are cached in context: ['before_agent_callback_state_key', 'before_model_callback_state_key', 'after_model_callback_state_key'] pass ✅
** Asserting keys are already persisted in session: ['before_agent_callback_state_key'] pass ✅
** Asserting keys are not persisted in session yet: ['before_model_callback_state_key', 'after_model_callback_state_key'] pass ✅
============================================================
[root_agent]: Hello! How can I help you verify something today?

===================== In after_agent_callback ==============================
** Asserting keys are cached in context: ['before_agent_callback_state_key', 'before_model_callback_state_key', 'after_model_callback_state_key', 'after_agent_callback_state_key'] pass ✅
** Asserting keys are already persisted in session: ['before_agent_callback_state_key', 'before_model_callback_state_key', 'after_model_callback_state_key'] pass ✅
** Asserting keys are not persisted in session yet: ['after_agent_callback_state_key'] pass ✅
============================================================
```

--------------------------------

### Configure and create a session with RedisSessionService

Source: https://github.com/google/adk-python/blob/main/docs/guides/integrations/redis/redis_session_service/index.md

Use RedisSessionServiceConfig to connect to a local Redis instance, then create a new session with initial state. The config sets the Redis URI, TTL, and key prefix.

```python
from google.adk.integrations.redis._redis_session_service import RedisSessionService
from google.adk.integrations.redis._config import RedisSessionServiceConfig

# Configure the service to connect to a local Redis instance
config = RedisSessionServiceConfig(
    uri="redis://localhost:6379",
    ttl_seconds=3600,
    key_prefix="myapp:"
)
service = RedisSessionService(config=config)

# Create a new session with initial state
session = await service.create_session(
    app_name="assistant_app",
    user_id="user_882",
    state={"user:name": "Alice", "step": "initial"}
)
```

--------------------------------

### Run agent.py to discover resource names

Source: https://github.com/google/adk-python/blob/main/contributing/samples/integrations/agent_registry_agent/README.md

Execute the sample script once to list registered agents and MCP servers to identify their full resource names.

```bash
python3 agent.py
```

--------------------------------

### Start remote prime agent server using adk api_server

Source: https://github.com/google/adk-python/blob/main/contributing/samples/a2a/a2a_basic/README.md

Starts the remote A2A server serving the check prime agent on port 8001.

```bash
# Start the remote a2a server that serves the check prime agent on port 8001
adk api_server --a2a --port 8001 contributing/samples/a2a/a2a_basic/remote_a2a
```

--------------------------------

### Navigate to the integration fixture directory

Source: https://github.com/google/adk-python/blob/main/tests/integration/fixture/bigquery_agent/README.md

Change to the parent directory containing the agent fixture folder before launching the web server.

```shell
cd tests/integration/fixture/
```
Tokens:
Raw