AgentAI

Why AgentAI?

Most agent frameworks are written in Python. AgentAI brings agents to Rust, with static typing, explicit error handling and a single binary to deploy. It is the foundation we use when building dedicated agents for our clients.

AgentAI uses the GenAI crate to talk to many LLM providers through one interface, so you can switch models without rewriting your agent.

Features

• Connect to any major LLM provider: OpenAI, Anthropic, Gemini, Ollama and other OpenAI-compatible APIs.
• Choose the right model for each step: pick a different model for each step of your agent’s workflow.
• Custom tools: a simple ToolBox interface and a #[toolbox] macro for exposing your own functions to the agent.
• MCP server support: reuse existing Model Context Protocol servers instead of writing tools from scratch.
• Structured output: describe a Rust type and get a typed response back, with no need to parse free-form text.

Quick start

cargo add agentai

use agentai::Agent;

#[tokio::main]
async fn main() -> anyhow::Result<()> {
    let mut agent = Agent::new("You are a useful assistant");
    let answer: String = agent.run("gpt-4o", "Why is the sky blue?", None).await?;
    println!("Answer: {}", answer);
    Ok(())
}

Roadmap

AgentAI is under active development and its API may still change. Planned next:

• Agent memory: control over persistence, record limits and context management.
• User input and streaming output: for interactive agents that need to talk to the user while they work.
• Configurable behavior: one place to configure model parameters and how tool errors are handled.