Speclib

You can run and connect to the Speclib MCP server to search, retrieve, and read SpecLib specifications through your MCP client. It exposes a concise set of tools to explore specs by query, scope, or content type and to fetch full specs or recipes, all via the Model Context Protocol.

How to use

You interact with the Speclib MCP server through your MCP client by connecting to either the HTTP endpoint or the local, stdio-based runtime. Use the provided tools to search specs, fetch full specs by ID or scope, list all scopes, or retrieve a bundled recipe. You can read a spec as markdown via the spec://{scope}/{slug} resource.

How to install

Prerequisites: you need Node.js version 20 or newer installed on your machine.

Step by step to get started locally:

1. Clone the repository for Speclib MCP.

2. Change to the project directory.

3. Install dependencies.

4. Run the MCP server locally.

Configuration and runtime details

The server expects an API URL for the SpecLib instance. By default, you use http://localhost:3000 as the SpecLib endpoint. You can override this with the SPECLIB_API_URL environment variable.

Notes and usage tips

You can use either an HTTP config to point at a running remote MCP endpoint or a local stdio configuration that runs the MCP server as a child process via npx. If you run the server locally, the recommended approach is to set SPECLIB_API_URL to your SpecLib instance URL and start the process as shown in the usage examples.

Available tools

search_specs

Search specs by query, scope, and/or content type to discover relevant specifications.

get_spec

Retrieve full specification content by ID or by scope/slug for in-depth reading.

list_scopes

List all available scopes to understand the organizational structure of specs.

get_recipe

Fetch a recipe together with its bundled specs for streamlined consumption.