Developer docs

MCP install

BacklinkMCP is a remote MCP server (JSON-RPC 2.0 over streamable HTTP) at https://backlinkmcp.com/api/mcp. No install, no local process — add the URL to your client's config.

Claude Desktop

Edit claude_desktop_config.json and restart the app.

{
  "mcpServers": {
    "backlinkmcp": {
      "url": "https://backlinkmcp.com/api/mcp"
    }
  }
}

Claude Code

From a project directory:

claude mcp add --transport http backlinkmcp https://backlinkmcp.com/api/mcp

Cursor

Add to .cursor/mcp.json (project) or the global MCP settings:

{
  "mcpServers": {
    "backlinkmcp": {
      "url": "https://backlinkmcp.com/api/mcp"
    }
  }
}

To use a Pro key over MCP, append it as a query param: https://backlinkmcp.com/api/mcp?key=YOUR_KEY. (MCP configs are a URL, so the query param is the norm here; for REST calls, prefer the Authorization header — see Auth.)

Recipes

The point of backlinks-over-MCP isn't a prettier dashboard — it's that an agent can chain the graph and the live web in one turn, then act on the result. Paste any of these into Claude (Desktop or Code) once the server is connected. Each is a workflow that would take several manual steps and a CSV export in a traditional tool.

recipeCompetitor backlink teardown → live outreach shortlist

Finds domains that link your competitors but not you (the gap you can close), then verifies which still publish outbound links today — so you pitch warm targets you don't already have, not stale ones.

Using backlinkmcp, run find_link_prospects with my domain mysite.com and
competitors stripe.com and squareup.com to get domains that link a competitor
but not me. Then run verify_link_live on the top prospects to confirm they
still publish outbound links today. Give me a table sorted by authority, and
for each, one line on why it might link to a payments product — as a backlink
outreach shortlist.

recipeVet a guest-post target before you pitch

Authority alone can be inflated by junk. This asks the agent to sanity-check the score against who actually links the site.

Before I pitch a guest post to smashingmagazine.com, use backlinkmcp to
get its domain_authority_score and its top linkers. Tell me whether the
authority looks earned (wiki/edu/news/real editorial links) or inflated by
low-quality or infrastructure links, and give me a go / no-go.

recipeDid my backlinks survive? (link reclamation)

Feed the agent the domains that used to link you; it checks each live and flags the drop-offs to chase.

Use backlinkmcp verify_link_live to check whether these domains still link
to my site example.com right now: theverge.com, techcrunch.com, wired.com.
List which ones are confirmed live and which dropped the link, so I can send
reclamation emails to the ones that dropped it.

verify_link_live only fetches homepages, so a "not_found" means the link isn't on the front page today — it may still live on an inner page. Treat drop-offs as leads to check, not confirmed removals.

recipeAudit your inbound links for spam

Scans your own linkers for spammy patterns — abused TLDs, spam keywords, link-farm outbound counts — and hands you a review list, not a trigger-happy disavow file.

Use backlinkmcp flag_toxic_links on my domain example.com. Group the flagged
domains by severity, and for the high-severity ones tell me which single signal
drove each flag. Remind me that Google says most sites should not disavow, and
help me decide which (if any) are worth a manual review.

flag_toxic_links is a heuristic on domain-level signals (TLD, name patterns, authority, outbound counts) — candidates to review, never a confirmed toxic link. Google's own guidance is that most sites should not use the disavow tool at all; only disavow with a manual action or clear negative-SEO.

recipeFind who links your whole niche

Intersects several competitors' backlink profiles to surface the publications, directories, and roundups that link the whole category — the ones you almost certainly need too.

Use backlinkmcp common_linkers on notion.so, airtable.com, and coda.io to
find the domains that link more than one of them. Prioritize the ones that link
all three — those are the must-have targets for a productivity tool — and give
me a shortlist with a one-line pitch angle for each.

MCP tools

Eight tools — three read from the cached graph engine, one fetches the live web, three run link-building analysis (competitor gap, toxic-link audit, and multi-domain overlap), and one (Pro) runs an AI-visibility panel across multiple LLMs. tools/list returns their full JSON Schema; here's the practical summary.

tooldomain_authority_score

Editorial-authority score for a domain, weighted toward wiki/edu/gov/news referrers.

Example structuredContent in the tool result:

{
  "domain": "stripe.com",
  "found": true,
  "domain_rating": 80.9,
  "referring_domains": 702082,
  "links_out": 41
}

tooltop_linkers_for_domain

Top 25 domains linking to a given domain, ranked by the linking domain's own authority.

toolbacklinks_for_domain

Full backlink payload: total inlink count plus the ranked linking domains — up to 100 rows on Pro, 25 on the free tier.

Example structuredContent (Pro key — the free tier returns row_cap 25):

{
  "domain": "stripe.com",
  "found": true,
  "total_inlinks": 702082,
  "row_cap": 100,
  "ranked_returned": 100,
  "results": [
    { "source": "en.wikipedia.org", "authority": 48213, "links_out": 3 },
    { "source": "github.com", "authority": 39187, "links_out": 12 }
  ]
}

toolverify_link_live

Live verification: fetches each linking domain's homepage right now and confirms an actual href to the target exists. Only homepages are fetched, so not_found means "no link on the homepage today", not "the link is gone".

Example structuredContent:

{
  "domain": "stripe.com",
  "checked": 8,
  "confirmed": 5,
  "results": [
    { "linker": "github.com", "status": "confirmed",
      "evidence": ["https://stripe.com/"], "checked_url": "https://github.com/" },
    { "linker": "en.wikipedia.org", "status": "not_found",
      "checked_url": "https://en.wikipedia.org/" }
  ],
  "note": "homepage-only check; not_found ≠ link removed"
}

toolfind_link_prospects

Competitor backlink-gap analysis: pass your domain and 1–3 competitors, get the domains that link a competitor but not you — ranked by the linking domain's authority. A warm link-building outreach shortlist. Domain-level only (the index has no per-link URLs or anchor text), so each row is a domain to pitch, not a confirmed link. CDN/analytics/marketing-widget/social noise is filtered out. Free mines 1 competitor and returns up to 25 prospects; Pro mines up to 3 and returns up to 100.

Example structuredContent (Pro, two competitors). authority is BacklinkMCP's relative link-authority score (higher = stronger linker), not a 0–100 rating; prospects linking multiple competitors are the warmest:

{
  "your_domain": "ugmonk.com",
  "your_linker_count": 50,
  "competitors": [
    { "domain": "studioneat.com", "ok": true, "total_inlinks": 445, "linkers_seen": 50 },
    { "domain": "baronfig.com", "ok": true, "total_inlinks": 358, "linkers_seen": 50 }
  ],
  "total_prospects_before_cap": 88,
  "filtered_junk_count": 2,
  "row_cap": 100,
  "prospects": [
    { "source": "kk.org", "authority": 4565, "links_out": 3291,
      "linked_competitors": ["studioneat.com", "baronfig.com"] },
    { "source": "uncrate.com", "authority": 1931, "links_out": 1628,
      "linked_competitors": ["studioneat.com"] }
  ]
}

toolflag_toxic_links

Toxic / low-quality backlink audit — the inverse of find_link_prospects. Scans a domain's own linkers and flags spammy disavow candidates: abused TLDs (.tk/.xyz/.loan…), spam keywords in the domain, hyphen/keyword stuffing, and link-farm-level outbound counts. Domain-level heuristic only (no per-link URL or anchor), so each row is a domain to review, not a confirmed toxic link. Note: Google says most sites should not disavow — use this to audit, not to auto-disavow. Free returns up to 25 flags, Pro up to 100.

Example structuredContent. toxicity_score is a 0–100 heuristic (higher = more signals); severity is high at 70+:

{
  "domain": "yoursite.com",
  "total_linkers_evaluated": 100,
  "total_flagged": 6,
  "high_severity": 2,
  "medium_severity": 4,
  "row_cap": 100,
  "flagged": [
    { "source": "cheap-viagra-pills.tk", "authority": 3, "links_out": 40,
      "toxicity_score": 110, "severity": "high",
      "reasons": ["high-abuse TLD (.tk)", "spam keyword in domain (\"viagra\")", "low authority (3) vs this profile"] },
    { "source": "linkdirectory.info", "authority": 6, "links_out": 9000,
      "toxicity_score": 50, "severity": "medium",
      "reasons": ["links out to 9000 domains — far above this profile's median (link farm/directory pattern)"] }
  ],
  "disclaimer": "Heuristic flags to review — Google says most sites should not disavow."
}

toolcommon_linkers

Multi-domain backlink overlap. Pass 2–4 domains, get the domains that link to more than one of them — ranked by overlap (how many of the set they link) then authority. Answers "who links my whole niche?" — the must-have publications, directories, and roundups your competitors all share. Domain-level only (no per-link URL or anchor). Free compares 2 domains, Pro up to 4; returns up to 25 rows free / 100 Pro.

Example structuredContent (Pro, three domains). overlap_count is how many of the set the linker links; domains linking all of them are the highest-priority targets:

{
  "targets": [
    { "domain": "studioneat.com", "ok": true, "linkers_seen": 50 },
    { "domain": "baronfig.com", "ok": true, "linkers_seen": 50 },
    { "domain": "ugmonk.com", "ok": true, "linkers_seen": 50 }
  ],
  "targets_resolved": 3,
  "total_overlap_before_cap": 12,
  "row_cap": 100,
  "common_linkers": [
    { "source": "kk.org", "authority": 4565, "links_out": 3291,
      "overlap_count": 3, "linked_targets": ["studioneat.com", "baronfig.com", "ugmonk.com"] },
    { "source": "uncrate.com", "authority": 1931, "links_out": 1628,
      "overlap_count": 2, "linked_targets": ["studioneat.com", "baronfig.com"] }
  ]
}

toolai_visibility_report Pro

AI Visibility / GEO audit. Asks a panel of AI models a buyer-intent question ("what's the best <category>?") and reports whether your brand gets recommended, who the models prefer instead (share-of-voice across the panel), why the leader wins — diagnosed against BacklinkMCP's own authority/link data — and which domains to earn links from to catch up. Two modes via the mode param: parametric (default — the models' baked-in brand memory, cheap) or retrieval (live web-search models = what the AI answers right now, SEO/citation-driven; costs more). Pro plan only (runs a live multi-model panel): 6 models × 3 prompts × up to 4 competitors per report, metered by a daily per-key credit budget.

Example structuredContent (Pro):

{
  "your_brand": "Airtable",
  "your_visibility": { "top_pick_rate": 0, "mention_rate": 0, "rank": null },
  "share_of_voice": [
    { "brand": "Trello", "top_pick_rate": 0.8, "mention_rate": 1.0 }
  ],
  "leader": { "brand": "Trello", "domain": "trello.com", "top_pick_rate": 0.8 },
  "diagnosis": {
    "your_referring_domains": 40884,
    "leader_referring_domains": 56608,
    "referring_domain_multiple": 1.4,
    "link_targets_to_close_gap": ["yelp.com", "dribbble.com", "booking.com"]
  }
}

REST endpoints

Same engine, plain HTTP GET. Good for scripts, backends, or anything that isn't an MCP client.

GET/api/v1/authority?domain=X

curl "https://backlinkmcp.com/api/v1/authority?domain=stripe.com"

{
  "ok": true,
  "authority": {
    "domain": "stripe.com",
    "found": true,
    "domain_rating": 80.9,
    "referring_domains": 702082,
    "links_out": 41
  }
}

GET/api/v1/backlinks?domain=X

curl "https://backlinkmcp.com/api/v1/backlinks?domain=stripe.com"

{
  "ok": true,
  "backlinks": {
    "domain": "stripe.com",
    "found": true,
    "total_inlinks": 702082,
    "ranked_returned": 100,
    "results": [
      { "source": "en.wikipedia.org", "authority": 98.2, "links_out": 3 },
      { "source": "github.com", "authority": 95.4, "links_out": 12 }
    ]
  }
}

GET/api/health

Pings the underlying engine. Useful for uptime checks or debugging a stuck integration.

curl "https://backlinkmcp.com/api/health"

POST/api/mcp

JSON-RPC 2.0 endpoint — this is what MCP clients call under the hood. You can hit it directly too:

curl -X POST "https://backlinkmcp.com/api/mcp" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"domain_authority_score","arguments":{"domain":"stripe.com"}}}'

Auth

The free tier needs no key at all — it's rate-limited by IP. Pro keys unlock the unlimited tier. Preferred: pass the key in the Authorization header so it never lands in URLs, server logs, or browser history:

curl "https://backlinkmcp.com/api/v1/authority?domain=stripe.com" \
  -H "Authorization: Bearer YOUR_KEY"

A ?key= query parameter also works as a fallback (handy for MCP clients that only take a URL), but avoid it where you can — query strings get logged:

curl "https://backlinkmcp.com/api/v1/authority?domain=stripe.com&key=YOUR_KEY"

The Authorization header wins if both are sent. Get a key on the Pro plan.

Rate limits

Free tier: 10 lookups/day, tracked per calendar day (UTC). Anonymous callers are counted by IP; a free API key is counted by key. Every REST call and every MCP tools/call counts against the same daily bucket. Pro keys are unlimited — that's the plan's core perk, alongside the full 100-row linker lists.

When the limit is hit, REST returns HTTP 429 with limited: true; MCP returns a normal tool result (not an error) with an upgrade message in the text content, so agents don't treat it as a crash.

Error shapes

Two distinct "not a success" shapes to check for, on both REST and MCP:

Rate limit reached

{
  "ok": false,
  "limited": true,
  "error": "Free daily limit reached (10/day). Upgrade: https://backlinkmcp.com#pricing"
}

Engine busy / upstream unavailable

{
  "ok": false,
  "busy": true,
  "error": "Engine temporarily unavailable, try again shortly."
}

Always check ok first. If ok is false, check limited to distinguish "you're out of free lookups" (retry tomorrow or upgrade) from busy (retry with backoff — transient).