Agent Development Kit (adk-rs)

An open-source, code-first Rust framework for building, evaluating, and deploying AI agents. adk-rs is a Rust port of Google's ADK: model-agnostic, deployment-agnostic, and wire-compatible with the Python ADK ecosystem (eval sets, adk api_server endpoints, the adk-web UI, and the A2A protocol) — with the low overhead, predictable latency, and safety guarantees of the Rust toolchain.

📖 Full documentation: adk-rs.vercel.app

Highlights

• Three providers, one trait — Gemini, Anthropic Claude, and any OpenAI-compatible endpoint (Azure, Ollama, Groq) behind Arc<dyn Model>, with native SSE streaming, image input, automatic retry/backoff, and prompt caching on every provider.
• Composable agents — LlmAgent, SequentialAgent, ParallelAgent, and LoopAgent nest through one BaseAgent trait, driven by a unified event stream, with automatic agent transfer between declared sub-agents and lifecycle callbacks around every agent, model, and tool call.
• Ergonomic tools — #[tool] on any async function derives the schema and wiring; or mount tools from any MCP server or OpenAPI 3.x spec.
• Production runtime — token-level streaming, structured output, extended thinking (Anthropic) with signature-safe round-trips, human-in-the-loop tool confirmation, pause/cancel/resume, context caching, event compaction, and semantic memory.
• Serve and interoperate — an axum dev server compatible with the adk-web UI, an embeddable CLI, spec-compliant A2A agent-to-agent JSON-RPC, and Gemini Live bidirectional audio streaming.
• Secure by default — HTTPS-or-loopback for credentials, redirect-stripped credential clients, loopback-only unauthenticated servers, Docker-sandboxed code execution, zero unsafe.

Installation

Default features are empty — opt in to what you need:

[dependencies]
adk-rs = { version = "0.6", features = ["gemini", "macros"] }
tokio = { version = "1", features = ["macros", "rt-multi-thread"] }
futures = "0.3"

Requires Rust 1.85+ (edition 2024). Each provider reads its key from the environment: GOOGLE_API_KEY, ANTHROPIC_API_KEY, or OPENAI_API_KEY (plus optional OPENAI_BASE_URL). See Installation for the full feature-flag reference.

Quick start

use adk_rs::agents::LlmAgent;
use adk_rs::providers::gemini::Gemini;
use adk_rs::runner::Runner;
use adk_rs::services::mem::InMemorySessionService;
use futures::StreamExt;
use std::sync::Arc;

#[tokio::main]
async fn main() -> adk_rs::Result<()> {
    let agent = LlmAgent::builder("greeter")
        .model(Arc::new(Gemini::from_env("gemini-2.5-flash")?))
        .instruction("You greet the user warmly.")
        .build()?;

    let runner = Runner::builder()
        .app_name("hello")
        .agent(Arc::new(agent))
        .session_service(Arc::new(InMemorySessionService::default()))
        .build()?;

    let mut events = runner.run("user", None, "Hello!").await?;
    while let Some(event) = events.next().await {
        if let Some(content) = event?.response.content {
            println!("{}", content.text_concat());
        }
    }
    Ok(())
}

Defining a tool

Annotate any async function — the macro derives the JSON schema and the Tool impl:

use adk_rs::tool;
use schemars::JsonSchema;
use serde::{Deserialize, Serialize};

#[derive(Deserialize, JsonSchema)]
struct GetWeatherArgs {
    /// City name in English (e.g. "Paris").
    city: String,
}

#[derive(Serialize)]
struct WeatherReport {
    temp_c: f32,
    description: String,
}

/// Look up the current weather in `args.city`.
#[tool]
async fn get_weather(
    args: GetWeatherArgs,
    _ctx: &mut adk_rs::core::ToolContext,
) -> adk_rs::Result<WeatherReport> {
    Ok(WeatherReport { temp_c: 22.0, description: "sunny".into() })
}

Attach it with .tool(get_weather()) on the agent builder.

Going further

Everything below is covered in depth at adk-rs.vercel.app:

Examples

Runnable demos live under examples/:

cargo run --example gemini_chat --features gemini
cargo run --example weather_agent --features "gemini,macros"
cargo run --example three_providers --features "gemini,anthropic,openai"
cargo run --example code_agent --features "code-exec,testing"
# live API wire-compatibility smoke test (needs real keys):
cargo run --example compat_check --features "anthropic,openai"

The documentation site itself lives under docs/ (cd docs && npm install && npm run dev); it is excluded from the published crate.

Contributing

Bug reports, feature requests, and pull requests are welcome. Before submitting:

cargo fmt --all
cargo clippy --all-features --all-targets
cargo test --all-features

Please open an issue before starting on substantial changes.

License

Licensed under the Apache License, Version 2.0 — see LICENSE.