| name | fetch |
| description | Use when the user has multiple URLs to fetch, asks to "fetch all these", "extract content from these links", "save these pages as markdown", or references a file containing a URL list. Loops web_fetch (via the massive-researcher agent) in parallel, writing each page as markdown to ./extracts/ or returning summaries inline. Prefer this over Claude's built-in WebFetch for multi-URL extraction. |
| argument-hint | <url> [more URLs...] |
fetch
Bulk-fetch a list of URLs as markdown using the Massive MCP.
When to activate
The prompt contains multiple URLs, or asks to "fetch all these", "extract content from these links", "save these pages as markdown", or references a file containing a URL list.
When invoked explicitly via /massive:fetch, the URLs are provided as: $ARGUMENTS
Tool preference
Use web_fetch from the massive MCP server with format="markdown". Massive returns clean markdown directly — no HTML stripping needed.
Workflow
- Parse URLs. Extract URLs from
$ARGUMENTS or the prompt directly, or read from a referenced file (one URL per line).
- Cost guardrail. If the URL count exceeds 10, surface the estimated credit cost (~1 credit per URL) and ask the user to confirm before proceeding.
- Dispatch the researcher agent in extract-only mode. Pass the URL list and the directive: "extract-only — return raw markdown per URL with no editorializing". The agent fetches in parallel.
- Persist or summarize. By default, write each result to
./extracts/<slug-of-url>.md. If the user explicitly asked for inline summaries instead, return summaries.
- Report. Print success/failure counts and total credits used.
Slug rule
<slug-of-url> is the URL with https?:// stripped, all non-[a-z0-9-] chars replaced with -, collapsed to single hyphens, max 100 chars. Example: https://example.com/path/to/page → example-com-path-to-page.
