Web Crawl

Web pages can be scraped one-off or crawled from a seed URL. Results land in the document search store and are queryable the same way as local PDFs and Office files.

--features crawl or --features full

Tools

All of the following are modes of the web MCP tool (basemind web <mode> on the CLI).

`web scrape`

Fetch and index a single page, adding it to the document store.

{
  "mode": "scrape",
  "url": "https://docs.example.com/guide/overview",
  "scope": "docs"
}

Results land tagged with a scope (default: web:<host>), searchable via memory mode documents alongside local documents. Override the scope to group related pages under a custom label.

`web crawl`

Follow links from a seed URL up to a configurable depth, indexing all discovered pages.

{
  "mode": "crawl",
  "url": "https://docs.example.com/",
  "max_depth": 3,
  "max_pages": 200
}

The crawler:

Honors robots.txt by default (disable with config).
SSRF-blocks private and loopback hosts by default to prevent exfil. Allow them with [crawl].allow_private_network = true in .basemind/basemind.toml.
Tags all pages under a scope of web:<host> by default, overridable per-request.
Stops when it hits max_depth, max_pages, or no new links.

`web map`

Discover a site’s pages without fetching bodies — returns a sitemap-like list.

{
  "mode": "map",
  "url": "https://docs.example.com"
}

Useful for reconnaissance: “what pages exist on this site?” without the download and embedding cost. Use web crawl if you want to index them all, or web scrape for one page.

Configuration

In .basemind/basemind.toml (config-file-only):

[crawl]
respect_robots_txt = true         # honor robots.txt (default: true)
allow_private_network = false     # allow crawling localhost, 10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16 (default: false)

These are config-file-only settings — the MCP tools do not expose overrides for security.

Security

The crawler SSRF-blocks private and loopback networks (localhost, 127.0.0.1, 10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16) to prevent exfiltration attacks. Set allow_private_network = true only if you trust the crawl target.

Searching crawled pages

Once indexed, crawled pages are searchable via memory mode documents:

{
  "mode": "documents",
  "query": "rate limiting",
  "mime_type": "text/html"
}

Filter by scope: "web:docs.example.com" to search only pages from a specific host, or search across all web pages with no filter.

Discipline

Use web scrape for one-off pages. Pull docs into context without committing to a full crawl.
Use web crawl for docs sites. Point it at /docs or /api and let it discover the rest.
Use web map for reconnaissance. Check what pages exist before committing to a crawl.
Search with mime_type: "text/html" to focus on web results. Keeps web docs separate from local files.
Expect crawls to take time. A 100-page site with embedding will take ~10–30 seconds.