A modern multi-agent framework written in Rust, AutoAgents enables developers to build, deploy, and coordinate intelligent agents powered by LLMs. It targets engineers and researchers creating complex AI systems, offering a performant, safe, and scalable foundation for cloud-native, edge, and browser-based applications. The framework's modular design and WASM compilation support are key benefits for flexible deployment.

How It Works

AutoAgents leverages Rust's performance and safety guarantees to provide a robust multi-agent system. Its core architecture is modular, allowing for swappable components like memory backends (currently supporting sliding window, with persistent storage planned) and executors (ReAct, basic, with streaming). It features provider-agnostic LLM integration, supporting numerous cloud and local models, and offers native WASM compilation for direct deployment in web browsers, enabling sandboxed tool execution via a WASM runtime.

Quick Start & Requirements

• Prerequisites: Rust (latest stable recommended), Cargo, LeftHook.
• Installation: Clone the repository, install Git hooks (lefthook install), and build the project (cargo build --release). The autoagents-cli crate provides a command-line interface.
• Primary Commands: cargo build --release for development. autoagents run --workflow <workflow.yaml> to execute workflows, and autoagents serve --workflow <workflow.yaml> to expose them via a REST API.
• Links: Documentation, Examples.

Highlighted Details

• Multi-Platform Deployment: Supports native Rust applications, WebAssembly for browsers, and edge deployments using ONNX models.
• Extensive LLM Provider Support: Integrates with major cloud providers (OpenAI, Anthropic, Google) and local solutions like Ollama.
• CLI & Serving: Includes an autoagents CLI for running and serving workflows defined in YAML, facilitating easy deployment and management.
• Type-Safe Tooling: Utilizes Rust's type system and derive macros for easy tool integration and structured output validation.

Maintenance & Community

The project is actively developed by the Liquidos AI team and community contributors. Community engagement is encouraged via GitHub Issues, Discussions, and a Discord server.

Licensing & Compatibility

AutoAgents is dual-licensed under the MIT License and Apache License 2.0, offering flexibility for commercial use and integration into closed-source projects.

Limitations & Caveats

Persistent memory storage is listed as "Coming Soon." Some LLM backends, such as Mistral-rs, Burn, and Onnx, are marked as experimental or under development, indicating potential instability or incomplete features.