Skip to main content

comment-scraper

❖ Communityv0.1.0

Tools · by Comment Scraper · added

Ask about the Reddit and Hacker News threads you exported with the Comment Scraper browser extension: list, read, search and quote them by comment number.

Open in Hermes Desktop
hermes plugins install comment-scraper
Repository ↗Reviewed source @ fd739bc ↗Documentation ↗

README

From the reviewed commit fd739bc ↗; it updates when the author re-pins.

Comment Scraper plugin

Ask about the Reddit and Hacker News threads you exported with the Comment Scraper browser extension, without pasting them in. The plugin gives the agent four read-only tools (list your exports, read a thread, search across them, show one comment in context) and a skill that tells it how to use them and cite comments by their [c12] numbers. The numbers match the extension's Markdown export, so you can find any quoted comment in your file.

What you need

  • The Comment Scraper browser extension, with at least one thread exported as JSON: "JSON, one row per comment" or "JSON, replies nested". CSV and Markdown exports aren't read.
  • Node.js 20 or newer.

Install

Claude Code

In Claude Code:

/plugin marketplace add supertj/comment-scraper-mcp
/plugin install comment-scraper@comment-scraper

Or from a terminal:

claude plugin install comment-scraper --marketplace supertj/comment-scraper-mcp

Hermes Agent

hermes plugins install comment-scraper
hermes plugins enable comment-scraper

If install can't find it in the catalog yet, install it straight from this repository:

hermes plugins install https://github.com/supertj/comment-scraper-mcp#plugin
hermes plugins enable comment-scraper

The tools show up from your next session.

What it runs

When the plugin starts, the app runs npx -y comment-scraper-mcp@0.1.0. The first time, npx downloads that exact version of the comment-scraper-mcp package from npm; after that it starts from the local cache. The server's source is in this repository. It runs as a local process while the app is open.

The server reads JSON exports in your Downloads folder and the folders directly inside it. It recognises them by their content, so renamed files are fine.

It doesn't write, move or delete files, and it makes no network requests. It doesn't run shell commands, store credentials or send telemetry. What the agent reads through it goes to your model provider, like the rest of the conversation. Exports don't contain usernames.

Reading another folder

The server takes its folders from --dir (repeatable) or from COMMENT_SCRAPER_DIRS (folders separated by :, or ; on Windows).

  • Claude Code: set COMMENT_SCRAPER_DIRS in the environment Claude Code starts from.
  • Hermes Agent: Hermes passes only a few basic variables such as PATH and HOME to MCP servers, so COMMENT_SCRAPER_DIRS from your shell won't reach it. Add a server with the same name to your Hermes config.yaml instead. It takes the place of the plugin's entry:
mcp_servers:
  comment-scraper:
    command: npx
    args: ["-y", "comment-scraper-mcp@0.1.0", "--dir", "~/Research/reddit"]

Things to ask

  • "List my exported threads."
  • "Read the CRM thread and group the complaints about pricing. Cite comment numbers."
  • "Search my exports for 'switched from' and summarise what people moved to."
  • "Show c41 in context."

Already added the server by hand?

If you added comment-scraper-mcp yourself before, remove that entry or the agent sees every tool twice. In Claude Code, run claude mcp remove comment-scraper. In Hermes, delete it from mcp_servers in config.yaml, unless it's named comment-scraper, in which case it simply replaces the plugin's own.

Support

Website: https://comment-scraper.com. Email: hello@comment-scraper.com. Privacy policy: https://comment-scraper.com/privacy

License

MIT

← Back to the catalog · catalog built Oct 17, 2026