Kage

Kage is a tool built to help AI coding agents remember what they learn about a software project. It works by storing important details like setup commands, known bugs, and coding conventions directly inside the code repository. This ensures that AI tools like Codex, Claude Code, and Cursor do not forget context when they return for future work. Without this tool, agents would have to ask the same questions and make the same mistakes every time they open a project. With Kage, they start with a clear understanding of the codebase.

Benefits

Kage offers several key advantages for developers and teams. It creates a shared memory that travels with the code, meaning every team member and AI agent starts from the same reviewed context. The memory is stored as plain JSON files that are visible in the git history, so users can inspect and trust the data. It does not require any external databases or API keys, keeping everything local and private. The system also filters out sensitive information like passwords or personal data before saving it. This makes it safe for teams to use without worrying about data leaks.

Use Cases

This tool is designed for teams that use AI coding assistants to write or fix code. Developers can use it to save specific commands they used to set up a project or to document workarounds for tricky bugs. When a new AI agent joins the project, it can read the stored memory packets to understand the project rules immediately. It is also useful for large organizations where different developers work on different parts of the same codebase. By using Kage, the team ensures that everyone follows the same conventions and avoids repeating past errors. It works best with projects that use Node.js and support the Model Context Protocol.

Pricing

Pricing details are not available in the provided information.

Vibes

Public reviews or testimonials are not available in the provided information.

Additional Information

Kage is available on npm with the latest version being v1.1.36. It requires Node.js version 18 or newer to run. The tool installs two main programs called kage and kage-graph-mcp. It supports popular AI agents such as Codex, Claude Code, and Cursor out of the box. Users can also configure it for any other client that supports the Model Context Protocol. The project follows a local-first approach, meaning all data stays on the user's computer and is not sent to a remote server.