Crate machi 

Machi - A Rust implementation of smolagents

This crate provides a lightweight, ergonomic framework for building AI agents that can use tools and interact with language models.

Quick Start

use machi::prelude::*;

#[tool(description = "Add two numbers")]
async fn add(a: i32, b: i32) -> Result<i32, ToolError> {
    Ok(a + b)
}

#[tokio::main]
async fn main() {
    // Use OpenAI (requires OPENAI_API_KEY env var)
    let model = OpenAIClient::from_env().completion_model("gpt-4o");

    // Or use Ollama (local, no API key needed)
    // let model = OllamaClient::new().completion_model("qwen3");

    let mut agent = Agent::builder()
        .model(model)
        .tool(Box::new(Add))
        .build();

    let result = agent.run("What is 2 + 3?").await;
}

Re-exports

pub use error::AgentError;

pub use error::Result;

Modules

agent

AI Agent for executing tasks with tools.

callback

Advanced callback system for agent events.

error

Error types for the machi framework.

managed

Managed agent system for multi-agent collaboration.

mcp

MCP (Model Context Protocol) client integration.

memory

Memory system for tracking agent steps and state.

message

Message types for agent-model communication.

multimodal

Agent types for multimodal data (images, audio).

prelude

Prelude module for convenient imports.

prompts

Prompt template system for AI agents.

providers

LLM Provider implementations for various model APIs.

tool

Tool trait and utilities for defining agent tools.

tools

Built-in tools for agents.

Attribute Macros

tool

A procedural macro that transforms a function into a machi::tool::Tool implementation.