agent-core

A Rust framework for building terminal UI agents powered by large language models.

agent-core provides the infrastructure to create production-ready LLM-powered terminal applications. It handles the complexity of managing async communication between UI components, LLM providers, and tool execution while offering flexible abstractions for customization.

Features

• Complete TUI components built on Ratatui (chat views, input widgets, permission panels)
• Session management with automatic context window compaction
• Tool execution framework with concurrent execution and permission handling
• Provider-agnostic LLM client (supports Anthropic and OpenAI)
• Customizable themes, layouts, and key bindings
• Slash command system for in-app commands
• Event-driven architecture with clean component separation

Getting Started

Add agent-core to your Cargo.toml:

[dependencies]
agent-core = "0.1.0"

Basic Usage

Create a minimal agent by implementing the AgentConfig trait:

use agent_core::agent::{AgentConfig, AgentCore};

struct MyAgent;

impl AgentConfig for MyAgent {
    fn config_path(&self) -> &str {
        "~/.config/myagent"
    }

    fn default_system_prompt(&self) -> &str {
        "You are a helpful assistant."
    }

    fn log_prefix(&self) -> &str {
        "myagent"
    }

    fn name(&self) -> &str {
        "MyAgent"
    }
}

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let agent = AgentCore::new(MyAgent)?;
    agent.run().await?;
    Ok(())
}

This creates a working terminal agent with chat UI, LLM integration, and tool execution capabilities.

Architecture

The framework has four main components:

• agent: Orchestrates all components and provides AgentCore for quick setup
• controller: Manages LLM sessions, tool execution, and event coordination
• client: HTTP client for LLM providers (Anthropic, OpenAI)
• tui: Terminal UI components (widgets, layouts, themes, commands)

Communication flows through async channels in an event-driven architecture. The controller coordinates between the UI, LLM provider, and tool execution layers.

Contributing

Contributions in any form (suggestions, bug reports, pull requests, and feedback) are welcome. If you've found a bug, you can submit an issue or email me at rsingh@arrsingh.com.

License

This project is dual-licensed under the MIT LICENSE or the Apache-2 LICENSE:

• Apache License, Version 2.0
• MIT License

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.

Contact: rsingh@arrsingh.com