Summary

This project offers a minimal, Git-based kernel to imbue any AI coding agent with statefulness. It enables agents to remember past interactions, maintain knowledge, and build upon previous work without requiring complex frameworks or databases, making state management exceptionally accessible for developers and power users.

How It Works

The kernel leverages an AI agent's inherent ability to process markdown files as instructions. Statefulness is achieved through a simple structure: AGENTS.md (generic kernel instructions), IDENTITY.md (agent persona), KNOWLEDGE.md (knowledge index), and a Git repository for persistent storage. The knowledge/ directory stores mutable facts, while notes/ logs session narratives append-only. This design bypasses the need for databases or vector stores, relying solely on plain markdown and Git for memory.

Quick Start & Requirements

• Primary Install: Clone the repository and run your preferred AI coding agent within the directory.

  git clone https://github.com/oguzbilgic/agent-kernel.git my-agent
  cd my-agent
  opencode # or claude, codex, cursor, etc.
  

• With kern-ai runtime:

  npx kern-ai init my-agent
  npx kern-ai tui
  

• Prerequisites: An AI coding agent (e.g., OpenCode, Claude Code, Codex, Cursor) and Git. kern-ai requires Node.js/npm.
• Links: GitHub Repository

Highlighted Details

• Enables statefulness for AI coding agents using only Git and markdown files.
• Eliminates the need for external databases, vector stores, or custom frameworks.
• Compatible with a range of AI coding agents including OpenCode, Claude Code, Codex, and Cursor.
• The optional kern-ai runtime provides daemon mode, Telegram/Slack integration, and unified cross-channel memory.
• Supports deploying multiple distinct agents, each managed in its own Git repository.

Maintenance & Community

No specific details regarding contributors, sponsorships, or community channels (e.g., Discord, Slack) are provided in the README.

Licensing & Compatibility

• License: MIT.
• Compatibility: Permissive for commercial use and integration with closed-source applications.

Limitations & Caveats

As a "minimal kernel," it may lack advanced features found in more comprehensive agent frameworks. The reliance on plain markdown and Git could present scalability challenges for extremely large knowledge bases or complex state management compared to dedicated solutions. The effectiveness hinges on the agent's capability to correctly interpret and manage the markdown memory files.