It seems like every other day I'm seeing new tools for scraping the web to Markdown. I would like to point out two of my own that seem to be constantly overlooked in the fray:

Gather: https://github.com/ttscoff/gather-cli — a solid command line tool that performs a custom Arc90 readability, advanced Markdown conversion (handling footnotes, tables, etc.) and integrates well as a skill in your AI assistant (working on an MCP for it)
Marky: https://fuckyeahmarkdown.com (a web-based tool with optional readability and an API for automation, ostensibly also usable as a skill)

GitHub - ttscoff/gather-cli

Contribute to ttscoff/gather-cli development by creating an account on GitHub.

GitHub
@ttscoff not sure if the MCP is worth it, MCP definitions have a context cost all the time and the models are super good at digging around a cli when they need it. Skills or even a AGENTS.md saying to use the tool with an example should be enough! MCPs these days are best when you need to handle auth stuff

@nuncamind @ttscoff yeah I’m with Drew. I just did that last night for a custom tool that chunks massive files and pre-processes them.

Although Claude‘s suggestion was to put the info about the tool and instructions for when and how to use it in the CLAUDE.md file of the project.

That seemed to work, but I haven’t investigated enough to learn when / how something like that should be in AGENTS.md

Regardless, there is no need for an MCP server for this.

@masukomi @ttscoff CLAUDE.md and AGENTS.md are the same idea, agents is just generic, codex and I think others use AGENTS