Internal Link Graph.

Operates on a finished crawl

Most "internal linking" SEO tools force you to re-crawl the site through their proprietary spider. crawlcrawl already stores the pages you crawled. internal-link-graph reuses that data — pass a crawl_id and the actor reconstructs the link graph from your pages table, runs PageRank (50-iteration cap, damping 0.85, dangling-mass redistribution), and union-find for weakly-connected components.

curl -X POST https://api.crawlcrawl.com/v1/actors/internal-link-graph \
  -H "Authorization: Bearer crk_..." \
  -d '{"crawl_id": 1234, "top_n": 50}'

# → 200
{
  "actor": "internal-link-graph",
  "data": {
    "stats": {
      "pages": 5000, "edges": 38214,
      "components": 1, "iterations": 23, "converged": true
    },
    "pagerank":       [ { "url": "...", "score": 0.014 }, ... ],
    "top_donors":     [ { "url": "...", "edges": 412 }, ... ],
    "top_recipients": [ { "url": "...", "edges": 187 }, ... ],
    "anchor_text_by_target": { "/pricing": ["pricing", "see prices", ...] },
    "orphan_clusters": [ { "size": 4, "urls": [...] } ]
  }
}

What's computed

• PageRank — power method with damping 0.85, dangling-mass redistribution, L1 convergence threshold 1e-6. Returns top-N scored URLs.
• Top donors — pages that link out the most. These are your hub pages.
• Top recipients — pages that receive the most internal links. These are what your site "votes for".
• Anchor text by target — for each high-rank page, the anchor text used to link to it across the site.
• Orphan clusters — weakly-connected components disconnected from the main graph. Pages here are essentially invisible to crawlers that link-follow.
• Stats — pages, edges, components, iterations to convergence.

When to use it

Internal-linking audits. The PageRank top-N tells you which pages get the most link equity from your own site. If your money page (pricing, lead form) isn't in the top 20, you have a structural problem — fix by adding links from hub pages.

Orphan detection. Pages that exist but aren't reachable from anywhere are SEO dead weight. The orphan clusters surface these in a single list.

Anchor-text consistency. The anchor-text-by-target map shows whether the site links to a target with consistent, descriptive anchors or with generic "click here" text — the latter dilutes signal.

Pre/post migration verification. Run before and after a site migration. The PageRank distribution and orphan list should be similar; large drift means routes broke.

Pricing

One page-credit per call (flat rate, regardless of crawl size — caps at 100,000 pages). The graph compute runs server-side with no chrome render or cloud cost. See full pricing →

Where it fits