Toka - A Modular LLM Agent Framework

Table of Contents

• Overview
• Core Features
• Quick Start

Overview

Toka is a highly extensible and secure Rust-based framework for building AI-powered agents. It provides a modular architecture that allows seamless integration of different AI models and functionalities. With Toka, users can create, export, and share agents with ease.

Key Highlights:

• Rust-powered security: Memory safety and robustness
• Pluggable agent system: Easily extendable with new functionalities
• Multi-LLM support (DeepSeek, OpenAI, Claude, Grok & more)
• Agent export and sharing: Save and load agents as .agent files
• Cross-platform compatibility (CLI chat, Twitter)
• Modular design for custom solutions

Core Features

Agent-Based Architecture

• Pluggable Agents: Define different types of AI agents with distinct behaviors
• Agent System: Create and manage multiple AI-powered agents
• Modular and pluggable: Extend functionalities dynamically
• Export and Import: Share agents across different environments

State & Memory Management

• Shared State System: Persist and retrieve agent memory

  • Context-aware conversations
  • Custom data injection for external knowledge
  • Multi-agent collaboration with shared knowledge

• Secure and Reliable:

  • Memory Safety: Built with Rust to prevent common memory vulnerabilities.

  • Sandboxed Execution: Ensures agents run in isolated environments.

  • Function Extensibility:

  • Easy to Add New Functions: The function system allows seamless extension.
  • Flexible Function Execution: Run functions with agent context and state awareness.

LLM Integration

• Multiple AI Provider Support:
  • OpenAI
  • GPT-4-Free
  • Claude
  • DeepSeek
  • Grok
• Customizable Model Selection
• Easy API Key Management

Platform Support

• CLI-based interaction
• Built-in Twitter agent
• Future extensibility for other platforms

API & Tooling

• Extensible Tool System:
  • Define and register custom tools
  • Function calling and execution support
  • Dynamic responses based on external data

Storage Layer

• Agent Serialization: Export and import agents in .agent format
• Stateful Execution: Persistent agent memory

Quick Start

1. Clone the repository:

git clone https://github.com/tokahub/Toka.git

1. Install dependencies:

cargo build

1. Examples and app expect G4F api running on Localhost:/1337

   3.1 Start G4F

    g4f api
   

   3.2 Run app (Full showcase)

   cargo run --bin app
   

   
   3.3 Run example

cargo run --example import_export

For more examples, check the app/examples/ directory in the repository.