what is claude-mega-brain fr?

claude-mega-brain is a plugin for Claude Code that gives the AI accurate knowledge about your specific project the moment a session starts, instead of letting it guess based on general training data. Normally, if you ask an AI assistant something like which database column stores a total, it might guess a common name that is wrong for your actual project. This tool fixes that by automatically loading your project's own documentation into the conversation from the start. It works by scanning your project for markdown files that include a small piece of structured information at the top, called frontmatter, describing what kind of document it is, such as a database table, a metric definition, or a service. When a session begins, it builds a compact summary of all these documented concepts and injects it automatically, so the AI already knows the exact schema, naming, and logic used in your project without needing to search for files or ask questions. The README includes a benchmark comparing this approach against having no context, against using Obsidian with a search plugin, and against just dumping raw markdown files into context. In their tests, this tool matched the best accuracy while using fewer tokens and responding faster, and it does this with zero extra tool calls, since the knowledge arrives already loaded rather than being fetched during the conversation. If no documented files are found in a project, the tool adds no meaningful delay. Setup is done through Claude Code's plugin system: you add the plugin marketplace entry, install the plugin, then run one setup command inside your project. Documentation files can live anywhere in the project, there is no dedicated folder requirement, and the format for writing them is a simple markdown file with a few standard fields like type, title, and description.