mcp-greasyfork-scripts
an mcp server that searches greasy fork and fetches userscript source code. gives your coding agent access to thousands of userscripts — tampermonkey, greasemonkey, violentmonkey scripts — as rich context.
what it does
two tools:
search-scripts
searches greasy fork by keyword. accepts up to 10 search queries in parallel, fetches all result pages concurrently (semaphore of 10), deduplicates by script id. returns results as a TSV table with id, name, description, install counts, ratings, dates, and code urls. maximum 1000 results displayed per call.
get-script-details
takes greasy fork script urls and fetches complete metadata plus the full javascript source code. returns a formatted markdown report with metadata headers and fenced code blocks.
how it works
- scrapes greasy fork search result pages (HTML parsing via regex on
data-script-*attributes) - fetches all pagination pages in parallel with concurrency control
- deduplicates scripts across multiple search queries
- fetches script metadata via greasy fork json api and source code from code urls
- no api keys or tokens required — greasy fork is fully public
install
add to your mcp config:
{
"mcpServers": {
"greasyfork": {
"command": "npx",
"args": ["-y", "mcp-greasyfork-scripts"]
}
}
}
tools
search-scripts
| parameter | type | required | description |
|---|---|---|---|
queries | string[] | yes | 1-10 search keywords, each searched independently |
returns TSV with columns: ID, Name, Description, Daily Installs, Total Installs, Rating, Created, Updated, Version, Link, Code URL
get-script-details
| parameter | type | required | description |
|---|---|---|---|
script_urls | string[] | yes | 1-10 greasy fork script urls |
returns markdown with metadata headers and full source code in fenced code blocks.
license
mit