Handing a model a Substack URL and asking it to fetch the page gets you navigation, subscription prompts and comment scaffolding along with the article. Separating title, author, subtitle and body is what makes a summary request land on the actual text. Public posts only — anything behind a paywall is out of scope.
A single-purpose server for reading Substack posts. Give it a post URL and it downloads and parses the page, returning the structured pieces rather than raw HTML.
- download_substack takes a Substack post URL and returns its content
- Title, author, subtitle and body are extracted separately rather than handed back as one blob
- Works with public Substack posts
Node.js v18 or higher. The package is mcp-substack, version 0.1.0, installed from a checkout with `npm install`; the client runs it by pointing node at lib/index.mjs. No account, no key.
Build from source — clone the repository and build it, then point your client at the binary