Br8n

What Branded-Mayhem-Collective-LLC/br8n does

Branded-Mayhem-Collective-LLC/br8n provides a small MCP interface for a local directory of Markdown documents. The directory acts as a portable knowledge base that an AI model can inspect without storing the content in a vendor-managed account. The repository includes a suggested brain layout with folders for working methods, decisions, exceptions, handoffs, and voice guidance.

The design assumes one file represents one subject and that documents record reasoning as well as conclusions. That structure gives a model material it can use when answering questions or identifying conflicts with earlier decisions. The server itself is a reader: the provided tools expose listing, file reading, and text search, with no documented tool for editing the files.

How it works

The server runs as a stdio MCP process. You provide the path to a brain directory when starting it, and an MCP-compatible client launches the process and communicates with it. The client can list available content, read a selected file, or search across the directory.

Search is based on the files themselves. Results identify the matching file and line location, allowing an agent to point back to the source used for an answer. There is no vector database, separate search index, user account, or hosted dependency described for the open repository. Moving the knowledge base means copying the directory to another machine and supplying its new path.

Setup and configuration

Create a brain directory from the repository template or use an existing Markdown directory. The template is located under template/brain/ and contains the documented category folders. Start the server with npx @br8n/mcp /path/to/brain, or install the package globally and run br8n /path/to/brain.

An MCP client configuration uses the npx command, the -y argument, the @br8n/mcp package, and the path to the brain directory. Claude Desktop, Claude Code, and Cursor are named as supported clients in the README. The material does not specify environment variables or additional configuration files.

Tools and capabilities

The Branded-Mayhem-Collective-LLC/br8n MCP server exposes three tools:

• brain_list lists the available brain content.
• brain_read reads a selected file.
• brain_search searches the Markdown files and returns matching file and line information.

Together, these tools support retrieval from an inspectable, plain-file knowledge base. The repository describes the format as Markdown-only and does not document semantic embeddings, vector retrieval, indexing, synchronization, or write operations.

Limitations and notes

The open implementation is local and stdio-based, so the client must be able to launch the process and access the brain directory. The README does not document an HTTP mode, authentication, hosted storage, permissions model, or automatic document maintenance. It also distinguishes this project from a different br8n package on PyPI.

The file layout and server are released under the MIT license. A hosted offering is mentioned separately, but the repository material does not establish that hosted features are part of this MCP package.