ADK-Rust: Build AI Agents in Rust with a Powerful Development Kit

Introduction

ADK-Rust is a production-ready Rust framework designed for building sophisticated AI agents. It provides a comprehensive and flexible toolkit, allowing developers to create AI agents with modular components for models, tools, memory, and real-time voice interactions. The framework is model-agnostic and deployment-agnostic, optimized for frontier AI models, and includes robust support for real-time voice agents.

The latest v2.1.0 release introduces significant enhancements, including portable, validated teams with exact handoff and delegation semantics, a Gemini Enterprise Agent Platform path, hardened ambient scheduling, runtime skill writing, and argument-level tool guardrails. It also features a new ADK-Rust-owned responsive runtime UI for conversations, team topology, event timelines, and shared state.

Why Use ADK-Rust

ADK-Rust stands out for its performance, modularity, and extensive feature set. It boasts sub-millisecond agent loop overhead, significantly outperforming other frameworks like LangGraph, making it ideal for high-performance applications. The framework is composed of 43 publishable crates, offering a tiered feature system that allows developers to include only the capabilities they need, from minimal agents to full enterprise solutions with real-time audio, code execution, and sandbox environments.

Key capabilities include an embedded runtime UI, powerful tool integration with #[tool] derives, support for MCP clients and servers, various workflow agents (sequential, parallel, loop, graph), portable multi-agent teams, coding agents, real-time voice and video, RAG, and advanced memory management. It also provides enterprise-grade features such as governed computer use, agentic commerce, security (guardrails, RBAC, SSO), and comprehensive observability with OpenTelemetry tracing.

Installation

Getting started with ADK-Rust is straightforward. First, install the cargo-adk subcommand:

cargo install cargo-adk

Then, you can scaffold a new OpenAI agent with an HTTP runtime and embedded UI using the quickstart command:

cargo adk new quickstart_agent --template api --provider openai
cd quickstart_agent
cp .env.example .env
# Open .env and replace the OPENAI_API_KEY placeholder, then:
cargo run

This will launch a server, and you can access the UI at http://127.0.0.1:8080/ui/ to interact with your agent.

Examples

ADK-Rust allows you to build a simple agent with minimal code. Here's an example of a basic Gemini agent:

use adk_rust::prelude::*;
use adk_rust::Launcher;

#[tokio::main]
async fn main() -> AnyhowResult<()> {
    dotenvy::dotenv().ok();
    let model = GeminiModel::new(&std::env::var("GOOGLE_API_KEY")?, "gemini-3.7-flash")?;

    let agent = LlmAgentBuilder::new("assistant")
        .instruction("You are a helpful assistant. Be concise and accurate.")
        .model(Arc::new(model))
        .build()?;

    Launcher::new(Arc::new(agent)).run().await?;
    Ok(())
}

The cargo-adk CLI also provides various templates and addons to scaffold projects for different use cases: