Tocket - Context Engineering Framework

What is Tocket?

Tocket is a context engineering framework for AI agents. It solves a critical problem: AI agents forget between sessions. Tocket gives them shared memory through context files with zero runtime pollution.

The Problem

AI agents and LLMs work best with shared context. Without it, each conversation starts from scratch. Tocket enables persistent context across sessions through simple YAML and Markdown files that any LLM can read.

How It Works

Tocket provides a pattern for organizing context files (.context/ directory) that agents can read. Each file serves a purpose: activeContext tracks current focus, systemPatterns document architecture, and productContext holds product knowledge. This approach works with Claude, Gemini, GPT, Cursor, and any LLM.

Key Features

Zero runtime pollution - context is just files, not code
Works with any LLM - Claude, Gemini, GPT, Cursor, etc.
Lightweight and portable - no heavy dependencies
Framework-agnostic - language and tool independent
Memory between sessions - persistent context for agents
Structured patterns - conventions for organizing knowledge

Compatible With

Tocket works with all major LLMs and development tools: Claude (Anthropic), Gemini (Google), GPT-4 (OpenAI), Claude Code (VS Code extension), Cursor IDE, and any LLM that can read files.

Get Started

Visit the GitHub repository at https://github.com/pedrocivita/tocket to learn more and start using Tocket for your AI agent projects.