Maiko lets you build concurrent Tokio applications without a single spawn or manual channel. Declare actors, declare subscriptions, and Maiko handles the routing. It is a topology-aware event router with actor lifecycle management for Tokio.

Actors are named components with their own state and lifecycle. They communicate through events routed by topic, without knowing who's listening. Think Kafka-style pub/sub, but embedded in your Tokio application.

Maiko comes with a built-in test harness for asserting on event flow, correlation tracking for tracing event propagation, and monitoring hooks for observability — all without external infrastructure.

Building concurrent Tokio applications often means manually creating, cloning, and wiring channels between tasks:

// Without Maiko: manual channel wiring
let (tx1, rx1) = mpsc::channel(32);
let (tx2, rx2) = mpsc::channel(32);
let (tx3, rx3) = mpsc::channel(32);
// Clone tx1 for task B, clone tx2 for task C, wire rx1 to...

// With Maiko: declare subscriptions, routing happens automatically
sup.add_actor("sensor",    |ctx| Sensor::new(ctx),    Subscribe::none())?;      // produces events
sup.add_actor("processor", |ctx| Processor::new(ctx), &[Topic::SensorData])?;   // handles sensor data
sup.add_actor("logger",    |ctx| Logger::new(ctx),    Subscribe::all())?;       // observes everything

Maiko sits between raw Tokio and full actor frameworks. Think of it as moving from breadboard to PCB — same components, reliable traces, testable as a whole.

Event-centric systems: processing stock ticks, device signals, telemetry pipelines, handling system events, data transformation.

Not ideal for request-response APIs or RPC patterns.

cargo add maiko

use maiko::*;

#[derive(Event, Clone, Debug)]
enum MyEvent {
    Hello(String),
}

struct Greeter;

impl Actor for Greeter {
    type Event = MyEvent;

    async fn handle_event(&mut self, envelope: &Envelope<Self::Event>) -> Result<()> {
        if let MyEvent::Hello(name) = envelope.event() {
            println!("Hello, {}!", name);
        }
        Ok(())
    }
}

#[tokio::main]
async fn main() -> Result<()> {
    let mut sup = Supervisor::<MyEvent>::default();
    sup.add_actor("greeter", |_ctx| Greeter, &[DefaultTopic])?;

    sup.start().await?;
    sup.send(MyEvent::Hello("World".into())).await?;
    sup.stop().await
}

See the examples/ directory:

• pingpong.rs - Event exchange between actors
• guesser.rs - Multi-actor game with topics and timing
• monitoring.rs - Observing event flow

cargo run --example pingpong
cargo run --example guesser

For detailed documentation, see Core Concepts.

Maiko includes a test harness (built on the monitoring API) for observing and asserting on event flow:

#[tokio::test]
async fn test_event_delivery() -> Result<()> {
    let mut sup = Supervisor::<MyEvent>::default();
    let producer = sup.add_actor("producer", |ctx| Producer::new(ctx), &[DefaultTopic])?;
    let consumer = sup.add_actor("consumer", |ctx| Consumer::new(ctx), &[DefaultTopic])?;

    let mut test = Harness::new(&mut sup).await;
    sup.start().await?;

    test.record().await;
    let id = test.send_as(&producer, MyEvent::Data(42)).await?;
    test.settle().await;

    assert!(test.event(id).was_delivered_to(&consumer));
    assert_eq!(1, test.actor(&consumer).events_received());

    sup.stop().await
}

Enable with features = ["test-harness"]. See Test Harness Documentation for details.

The monitoring API provides hooks into the event lifecycle - useful for debugging, metrics, and logging:

use maiko::monitoring::Monitor;

struct EventLogger;

impl<E: Event, T: Topic<E>> Monitor<E, T> for EventLogger {
    fn on_event_handled(&self, envelope: &Envelope<E>, topic: &T, receiver: &ActorId) {
        println!("[handled] {} by {}", envelope.id(), receiver.name());
    }
}

let handle = sup.monitors().add(EventLogger).await;

Enable with features = ["monitoring"]. See Monitoring Documentation for details.

• Core Concepts - Events, Topics, Actors, Context, Supervisor
• Monitoring - Event lifecycle hooks, metrics collection
• Test Harness - Recording, spies, queries, assertions
• Advanced Topics - Error handling, configuration, design philosophy
• FAQ - Common questions answered
• API Reference - Complete API documentation

Near-term:

• Dynamic actor spawning/removal at runtime
• Improved supervision and error handling strategies

Future:

• maiko-actors crate with reusable actors (IPC bridges, WebSocket, OpenTelemetry)
• Cross-process communication via bridge actors

Maiko powers the daemon in Charon - a USB keyboard pass-through device built on Raspberry Pi. The daemon uses Maiko actors to read input from multiple keyboards, map and translate key events, output USB HID to the host, and coordinate telemetry, IPC, and power management.

Maiko is battle-tested in the Charon project, where it runs continuously, but it's not yet production-grade. I'd describe it as solid for happy-path scenarios and still maturing for rainy days. Flow control with per-topic overflow policies is in place, supervision is minimal, and improved error handling and recovery strategies are planned.

For now, Maiko demonstrates what it wants to be. That's the state I wanted to reach before sharing it with a wider audience. Want to help shape what comes next? See below.

Contributions welcome! Whether it's a bug report, feature idea, or pull request - all input is appreciated.

• Try it out and let us know what you think
• Report issues via GitHub Issues
• Looking to contribute code? Check out good first issues

Inspired by Kafka (topic-based routing) and built on Tokio (async runtime).

Licensed under the MIT License.