AgentAI

AgentAI is a Rust library designed to simplify the creation of AI agents. It leverages the GenAI library to interface with a wide range of popular Large Language Models (LLMs), making it versatile and powerful. Written in Rust, AgentAI benefits from strong static typing and robust error handling, ensuring reliable and maintainable code. Whether you're developing simple or complex AI agents, AgentAI provides a streamlined and efficient development process.

Warning

This library is under heavy development. The interface may change at any time without notice.

Features

Connect to any major LLM provider: Support for OpenAI, Anthropic, Gemini, Ollama, and other OpenAI-compatible APIs.
Choose the right model for the job: Flexibly select the best-suited model for each step in your agent's workflow.
Build custom tools with ease: A simple interface for creating and managing your own tools using the ToolBox.
MCP Server Support: Leverage existing solutions based on the Model-Context-Protocol, eliminating the need to build agent tools from scratch.
Structured Output: No need to parse raw text from model, just provide structure, and AI agent will provide response in defined format.

What's New

genai 0.7 (version 0.2.0)

AgentAI now uses genai 0.7, schemars 1.x and Rust edition 2024 (MSRV 1.88). Models that return text together with tool calls now work: the tool calls are no longer dropped. Agent::new and Agent::new_with_url now return Result, and a tool call without a toolbox is answered with an error message to the model instead of a panic. McpToolBox now uses the official MCP SDK (rmcp) and can also connect to running servers over Streamable HTTP. Agent::new_with_provider binds an agent to any genai provider, for example the native OpenRouter adapter (AdapterKind::OpenRouter), with an explicit API key. Agent::with_max_iterations and Agent::with_chat_options configure the loop limit and the model options (temperature, max tokens, tool choice, reasoning effort, ...). The genai crate is re-exported as agentai::genai. #[tool(strict)] marks a tool definition as strict, so supporting providers make the model's arguments match the parameter schema exactly.

`ToolBox` (version 0.1.5)

This release introduces the ToolBox, a new feature providing an easy-to-use interface for supplying tools to AI agents.

Future Plans

We are continuously working to improve AgentAI. Here are some of the features planned for the near future:

Agent Memory: Enhance the user experience by adding new functionality for AI agent memory. This will give users control over memory behavior, such as persistence, record limits, and context management for new requests.
User Input and Streaming Output: Not every AI agent works silently in the background. Some require additional interaction from the user. This feature will introduce an interface to handle user interactions and provide responses in a streaming format for a better user experience.
Configurable Behavior: Introduce a comprehensive way to manage every aspect of an agent's configuration, from model parameters to error-handling behavior for tools.

Installation

To add the AgentAI crate to your project, run the following command in your project's root directory:

cargo add agentai

This command adds the crate and its dependencies to your project.

Feature Flags

Available features for agentai crate. To enable any of these features, you need to enter this command:

cargo add agentai -F mcp-client

Features list:

mcp-client (enabled by default) — Enables support for Agent Tools served by MCP servers (stdio and Streamable HTTP, via rmcp)
macros (enabled by default) — Enables support for macro #[toolbox]
tools-buildin (enabled by default) — Enables support for buildin tools
tools-web (enabled by default) — Enables support for web tools

Usage

Here is a basic example of how to create an AI agent using 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(())
}

Error handling

Agent::run returns an AgentError that you can match on. Provider failures that may succeed on a second try report it with AgentError::is_transient:

use agentai::{Agent, AgentError};

match agent.run::<String>("gpt-4o", "Why is the sky blue?", None).await {
    Ok(answer) => println!("Answer: {answer}"),
    Err(err) if err.is_transient() => println!("Try again later: {err}"),
    Err(AgentError::InvalidAnswer { raw, .. }) => println!("Unexpected answer: {raw}"),
    Err(err) => println!("Failed: {err}"),
}

Tools report failures with ToolError. Recoverable errors (NotFound, InvalidArguments, Failed) are sent to the model as the tool result and the run continues. Fatal errors (Configuration, Fatal) stop the run with AgentError::Tool. ToolError::is_recoverable tells the two groups apart.

Examples

For more examples, check out the examples directory. To run an example, use the following command, replacing <example_name> with the name of the example file (without the .rs extension):

cargo run --example <example_name>

For instance, to run the simple example:

cargo run --example simple

Documentation

Full documentation is available on docs.rs.

Contributing

Contributions are welcome! Please see our CONTRIBUTING.md for more details.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgements

Special thanks to the creators of the GenAI library for providing a robust framework for interfacing with various LLMs.

agentai 0.2.0