> ## Documentation Index
> Fetch the complete documentation index at: https://docs.neosantara.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# Google ADK

> Build and coordinate autonomous AI agents using Google Agent Development Kit (ADK) and Neosantara via LiteLLM.

[Google Agent Development Kit (ADK)](https://adk.dev/?utm_source=neosantara-docs\&utm_medium=referral) is an open-source, code-first framework designed for building, evaluating, and deploying autonomous AI agents. ADK includes a built-in `LiteLlm` model wrapper, enabling ADK agents to execute any Neosantara model using the `neosantara/<model>` prefix.

## Setup

Install the Google ADK and LiteLLM packages:

```bash theme={"theme":{"light":"ayu-dark","dark":"catppuccin-latte"}}
pip install -U google-adk litellm
```

<CodeGroup>
  ```bash Bash / zsh icon="terminal" theme={"theme":{"light":"ayu-dark","dark":"catppuccin-latte"}}
  export NEOSANTARA_API_KEY="nsk_your_api_key_here"
  ```

  ```env .env icon="file-code" theme={"theme":{"light":"ayu-dark","dark":"catppuccin-latte"}}
  NEOSANTARA_API_KEY=nsk_your_api_key_here
  ```

  ```powershell PowerShell icon="terminal" theme={"theme":{"light":"ayu-dark","dark":"catppuccin-latte"}}
  $env:NEOSANTARA_API_KEY="nsk_your_api_key_here"
  ```
</CodeGroup>

## Basic Agent Initialization

Use the `google.adk.models.lite_llm.LiteLlm` class with the `neosantara/` provider prefix to connect Neosantara models to your ADK agent. Credentials are automatically loaded from the `NEOSANTARA_API_KEY` environment variable.

```python theme={"theme":{"light":"ayu-dark","dark":"catppuccin-latte"}}
import os
import asyncio
from google.adk.agents import Agent
from google.adk.models.lite_llm import LiteLlm
from google.adk.runners import InMemoryRunner

# Initialize an agent with a Neosantara model
researcher = Agent(
    name="researcher_agent",
    model=LiteLlm(model="neosantara/gemini-3.8-flash"),
    instruction="You are a technology research assistant that presents data concisely and objectively."
)

runner = InMemoryRunner(agent=researcher)

async def main():
    events = await runner.run_debug(
        "Explain the benefits of a regional AI gateway for developers in Indonesia.",
        quiet=True
    )
    for event in events:
        if event.is_final_response() and event.content:
            for part in event.content.parts:
                if hasattr(part, "text") and part.text:
                    print(part.text)

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

## Function and Tool Calling

Google ADK supports standard Python functions as agent tools. Pass `allowed_openai_params=["tools"]` to your `LiteLlm` instance to propagate tool schema declarations:

```python theme={"theme":{"light":"ayu-dark","dark":"catppuccin-latte"}}
import os
import asyncio
from google.adk.agents import Agent
from google.adk.models.lite_llm import LiteLlm
from google.adk.runners import InMemoryRunner

def calculate_token_estimate(text: str) -> str:
    """Calculate token and compute estimation based on character length."""
    estimated_tokens = len(text) // 4
    return f"Text length: {len(text)} characters. Estimated tokens: {estimated_tokens}."

analyst = Agent(
    name="token_analyst",
    model=LiteLlm(
        model="neosantara/gemini-3.8-flash",
        allowed_openai_params=["tools"]
    ),
    instruction="Use calculate_token_estimate when the user asks about text payload sizing.",
    tools=[calculate_token_estimate]
)

runner = InMemoryRunner(agent=analyst)

async def main():
    events = await runner.run_debug(
        "What is the token estimate for a 1000-character payload?",
        quiet=True
    )
    for event in events:
        if event.is_final_response() and event.content:
            for part in event.content.parts:
                if hasattr(part, "text") and part.text:
                    print(part.text)

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

## Multi-Agent Hierarchies (Sub-Agents)

ADK coordinates multi-agent workflows using the `sub_agents` parameter. A root agent orchestrates and delegates specialized queries to sub-agents:

```python theme={"theme":{"light":"ayu-dark","dark":"catppuccin-latte"}}
import os
import asyncio
from google.adk.agents import Agent
from google.adk.models.lite_llm import LiteLlm
from google.adk.runners import InMemoryRunner

# Specialized analytical reasoning sub-agent
analyst_agent = Agent(
    name="analyst",
    model=LiteLlm(model="neosantara/deepseek-v4.1-flash"),
    instruction="Perform rigorous quantitative reasoning and throughput latency analysis."
)

# Root orchestrator delegating domain queries
lead_orchestrator = Agent(
    name="coordinator",
    model=LiteLlm(model="neosantara/gemini-3.8-flash"),
    instruction="Coordinate technical questions. Delegate in-depth calculations to the analyst sub-agent.",
    sub_agents=[analyst_agent]
)

runner = InMemoryRunner(agent=lead_orchestrator)

async def main():
    events = await runner.run_debug(
        "Calculate concurrency efficiency for 1000 RPM on the Basic tier.",
        quiet=True
    )
    for event in events:
        if event.is_final_response() and event.content:
            for part in event.content.parts:
                if hasattr(part, "text") and part.text:
                    print(part.text)

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

## Next Steps

* [Model Context Protocol (MCP) Guide](/en/agents/overview)
* [Model Catalog & Pricing](/en/gateway/models)
* [LiteLLM Native Integration](/en/integrations/litellm)
* [CrewAI Multi-Agent Integration](/en/integrations/crewai)


## Related topics

- [LiteLLM](/en/integrations/litellm.md)
- [CrewAI](/en/integrations/crewai.md)
- [Agno](/en/integrations/agno.md)
- [Model Catalog](/en/gateway/models.md)