Is the Backlink MCP hosted like SellOnLLM’s GA4 MCP?

No. The Backlink MCP is open source and runs on your machine (stdio). SellOnLLM’s hosted connectors for Google Analytics and Search Console use HTTPS URLs such as https://www.sellonllm.com/api/mcp—that is a different product surface.

Do I need paid SEO APIs for the Backlink MCP?

No. The server uses DuckDuckGo search, the Wayback Machine CDX API, and HTTP scraping with httpx and BeautifulSoup. Results are samples, not a full commercial link index.

How do I install the Backlink MCP for Claude Desktop?

Clone https://github.com/vipul510-web/mcp-backlink-for-seo, create a Python virtual environment, install dependencies from the README, then add an mcpServers entry in claude_desktop_config.json pointing command to your venv Python and args to server.py. Restart Claude Desktop.

Open source · Python · stdio MCP

Backlink MCP for Claude & Cursor

Automate backlink research, unlinked mention hunting, and outreach prep inside your AI assistant. This connector is an MIT-licensed Python MCP on GitHub—it runs locally (no SellOnLLM OAuth URL). Pair it with our hosted GA + GSC MCP when you want rankings and traffic alongside link prospects.

Local stdio server — not the same as https://www.sellonllm.com/api/mcp

Why a Backlink MCP?

Classic SEO suites charge heavily for link indexes. This MCP gives individuals and agencies a free, automatable path to: surface pages that mention a brand (linked or not), narrow guest-post and resource-page angles, see who links to competitors, verify whether a page links to you, and pull contact signals for outreach—all orchestrated by Claude or Cursor through typed tools instead of brittle copy-paste.

Tools (from the MCP)

find_mentions — Discover pages mentioning your domain (linked or unlinked).
find_prospects — Guest posts, resource pages, and roundups by niche.
find_competitor_link_sources — Pages linking to a competitor (outreach targets).
verify_page_links — Scrape a URL to confirm links to you and gather on-page hints.
extract_contact_info — Emails, socials, and contact pages from a site.
check_page_history — Wayback checks to see if a URL has a stable history.

Signals come from DuckDuckGo, Wayback CDX, and direct fetches—use judgement on rate limits, robots directives, and recipient privacy (GDPR, CAN-SPAM, etc.).

Quick install

Clone vipul510-web/mcp-backlink-for-seo and create a Python 3.10+ venv.
Install: pip install mcp duckduckgo-search httpx beautifulsoup4 lxml (see repo for pinned versions).
Register server.py in Claude Desktop or .cursor/mcp.json with your absolute Python path.

Example Claude Desktop snippet (adjust paths):

{ "mcpServers": { "backlink-mcp": { "command": "/absolute/path/to/mcp-backlink-for-seo/.venv/bin/python", "args": ["/absolute/path/to/mcp-backlink-for-seo/server.py"] } } }

Open README

Typical automation workflow

Ask your assistant to chain tools: start from find_prospects or find_competitor_link_sources, filter to realistic domains, run verify_page_links on finalists, then extract_contact_info before drafting outreach. For brand monitoring, begin with find_mentions and split linked vs unlinked follow-ups.

Pair with SellOnLLM analytics

Backlinks move rankings over time—validate impact in GA + GSC chat or the Claude MCP for SEO (referring URLs, landing pages, query deltas). For AI answer surfaces, add the AI visibility MCP.

FAQ

Is this connector hosted on sellonllm.com?

No. It is a stdio MCP you run from source. Hosted SellOnLLM MCPs use HTTPS endpoints documented on the GA+GSC and AI visibility pages.

Does it replace Ahrefs or Semrush?

It does not provide a full-site link graph. It is best for research and prioritization without paid API keys—expect sampled SERP-style results, not enterprise-scale crawls.

Smithery or other registries?

The upstream repo may ship a smithery.yaml; you can also publish the Git URL per your registry’s instructions. This page is the canonical marketing URL on SellOnLLM.