About wyrd
Wyrd is a collection of read-only MCP servers and a containment library for exposing a user-approved folder. It is intended for AI clients that need folder access bounded by a documented path boundary.
Key features
- 01Read-only access to a granted folder
- 02Path-boundary containment library
- 03Separate MCP server packages
Use cases
- 01Let an AI client inspect a selected folder
- 02Build or inspect the server packages
MCP support
Wyrd provides MCP servers that expose one explicitly granted folder to clients, with path containment applied.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y wyrd-mcpWYRD_GRANTAbsolute path to the single folder to serve, read-only. Used when --grant is not passed. The server refuses to start if neither is set.WYRD_SEARCH_BACKENDOptional. Absolute path to one module file that answers search instead of the built-in search. Used when --search-backend is not passed. The module runs inside the server process and is not confined to the granted folder; leave unset unless you trust it.Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEWyrd
Wyrd is a collection of MCP servers and their containment library. Each server operates over one folder explicitly granted by the user and applies the library's documented path boundary.
Install the Reader
The published npm package is wyrd-mcp. For Claude Code, run:
claude mcp add wyrd -- npx -y wyrd-mcp --grant "<absolute folder>"
See the Reader README for other clients, the security boundary, and known limits. AI agents should follow llms-install.md.
Build from source
Build the workspaces from a clean checkout with:
npm install --ignore-scripts --package-lock=false
npm run build
Each package's README documents that package's usage, behavior, security boundary, and test
surface. Its package-local LICENSE defines its license. The repository scaffold itself is MIT.
Published npm artifacts remain the supported installation path; this repository exposes source and
configuration for inspection and building, with dependencies resolving inside their declared
ranges.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.