Linklight is an MCP server that lets AI agents perform SEO and link-building tasks: finding prospects, researching domains, drafting outreach emails, and monitoring backlinks/keywords.
Find prospects: search_prospects(keyword) returns roundup/list/resource-page targets; find_similar_prospects(url) returns semantically similar sites via Exa neural search.
Competitor research: find_competitor_backlinks(domain) finds pages linking to competitors and can filter out domains that already link to your site.
Domain enrichment: enrich_domain(domain) gives DA, contact email, title/description; find_email(domain) looks up contacts via Hunter.
Draft and save outreach: draft_email(topic) writes personalized emails (with optional prospect URL/name/tone) and returns a spam score; save_draft(prospect_id, subject, body) stores drafts for manual review—sending is never automated.
Campaign/prospect management: list_campaigns, list_prospects, list_replies let you query your outreach pipeline and reply activity.
Backlink monitoring: list_backlinks and list_lost_backlinks show healthy or broken/redirected/unreachable backlinks, optionally filtered by time.
Keyword opportunities: find_quick_win_keywords(site_id) surfaces striking-distance keywords (positions 11–30) from Search Console.
Outreach prioritization: find_prospect_gaps(campaign_id) lists prospects missing emails, sorted by DA, for efficient follow-up.
Provides email drafting capabilities by leveraging OpenAI's language models to generate personalized outreach emails.
The MCP server for SEO. Plug linklight into Claude Desktop, Claude Code, Cursor, or any MCP client and let your AI agent find prospects, draft outreach, and monitor backlinks — you approve and send.
Dashboard included. But the agent is the point.
Sign in at lightlinks.dev with Google.
Go to Settings → API Access → Generate key. Copy it.
Paste this into ~/.claude/settings.json (or your MCP client's config):
{
"mcpServers": {
"linklight": {
"type": "http",
"url": "https://lightlinks.dev/api/mcp",
"headers": { "Authorization": "Bearer sk_ll_PASTE_YOUR_KEY_HERE" }
}
}
}Restart your MCP client. Ask your agent: "List my linklight campaigns."
Full docs: lightlinks.dev/docs/mcp.
Related MCP server: seo-mcp
Tool | What it does |
| Find prospect sites for a topic (Tavily-backed, DA-enriched) |
| Find pages linking to a competitor in roundups/alternatives lists — pass |
| Given one good prospect, return 20 more like it (Exa.ai neural search) |
| DA + contact email + homepage title/description |
| Look up a contact email (Hunter) |
| Write a personalised outreach email with a built-in spam score |
| Save a drafted email against a prospect for you to review |
| Return striking-distance keywords (pages 2-3 with impressions) |
| Prospects in a campaign missing an email, sorted by DA |
| Backlinks currently broken, unreachable, or redirected |
| Query your data |
Sending emails is never exposed as an MCP tool. Every send needs your manual approval in the dashboard.
Find the top 20 prospects for "nextjs seo" with DA ≥ 40.
For each, draft a warm personalised email referencing their most recent post,
save each draft against the prospect, and show me the spam scores.Who links to ahrefs.com in roundups or "best of" articles?
Exclude domains that already link to my site, and give me the top 15 by DA.List my lost backlinks from the past 30 days, group by domain,
and tell me which ones are worth reaching out to.For campaign "Q4 outreach", find prospects without emails and run find_email
on the top 10 by domain authority.$19/mo. 7-day free trial, no credit card required.
Next.js 16 (Turbopack) on Vercel
Supabase (Postgres) with RLS on every table
NextAuth + Google OAuth (Gmail + Search Console)
Tavily for real-time web search
Exa.ai for semantic prospect discovery
OpenAI for email drafting
Moz + Hunter for enrichment
See docs/mcp/directory-submission.md for ready-to-paste listing details.
MIT — see LICENSE.
Generate an outreach email draft. Returns subject, HTML body, plain-text body, and a spam score (0-10, higher = better). Pass prospect_url to fetch the target's page and personalize the draft with a reference to their actual content — this dramatically improves reply rate over generic drafts.
Inputs: topic (required), tone, site_name, prospect_url, +3 more.
Return known facts about a domain: Moz Domain Authority, cached contact email, homepage title/description. Data is shared across all users of linklight so common domains are instant.
Inputs: domain (required).
Return pages that link to or feature a competitor in roundups / lists / alternatives / vs. articles — the classic 'who could realistically link to me if they link to my competitor' list. Optionally excludes domains that already link to my_domain (pass my_domain to get only NEW opportunities). Returns url, title, domain, position, and Moz Domain Authority.
Inputs: competitor_domain (required), limit, my_domain.
Look up a contact email for a domain via Hunter. Cached in domain_facts on hit.
Inputs: domain (required).
Return prospects in a campaign that are missing a contact email, sorted by Domain Authority DESC. Use to answer 'which prospects should I run find_email on next?'
Inputs: campaign_id (required), limit.
Return keywords the caller's site is ranking for in Search Console, filtered to 'quick win' opportunities: position 11-30 with meaningful impressions. Sorted by opportunity score (impressions ÷ position). Use to answer 'what should I write about next?'
Inputs: site_id (required), limit, position_max, position_min, +1 more.
Given a known-good prospect URL, return semantically similar URLs via Exa.ai's neural search. Use this when the caller already has one great prospect and wants 5-20 more like it. Excludes the source domain by default.
Inputs: url (required), limit.
List backlinks earned to a site. Filter by health_status (healthy|redirected|broken|unreachable|pending|error).
Inputs: limit, site_id, health_status.
List the caller's campaigns with id, name, status, and created_at.
Inputs: no named parameters.
Return backlinks in unhealthy states (broken, unreachable, redirected). Use to answer 'what did I lose recently?' Combine with a since filter to scope to a time window.
Inputs: limit, since, site_id.
List prospects. Filter by campaign_id and/or status (prospect|contacted|replied|live_link|declined|archived).
Inputs: limit, status, campaign_id.
List prospects who replied to outreach. Optionally filter by ISO-8601 since date.
Inputs: limit, since.
Save a drafted email as a note on a prospect. Does NOT send — user must review and send from the linklight UI.
Inputs: subject (required), body_html (required), prospect_id (required), body_text.
Find link-building prospect sites for a keyword. Prefers roundup / list / resource-page targets over direct competitor product pages. Uses the shared SERP cache when fresh; hits Tavily on miss. Returns url, title, domain, position, and Moz Domain Authority.
Inputs: keyword (required), limit.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 14 tools
Most tools have clearly distinct purposes: discovery (search, similar, competitor), enrichment (domain, email), listing (prospects, backlinks, replies, campaigns), outreach (draft, save), and analysis (quick win keywords, gaps). Minor overlap exists (list_backlinks vs list_lost_backlinks, enrich_domain vs find_email) but descriptions clarify the intended use cases.
All tool names use snake_case with a verb_noun structure, consistently using 'list' for listing actions and 'find' for discovery. The only deviation is 'search_prospects' using 'search' instead of 'find', but this is acceptable given the keyword-based nature of the search.
With 14 tools, the set is well-scoped for a link-building workflow, covering discovery, enrichment, outreach, and monitoring. Each tool serves a distinct purpose and none are redundant or superfluous, fitting the 3-15 tool sweet spot.
The toolset covers discovery, enrichment, outreach drafting, and backlink monitoring, but lacks lifecycle management operations such as updating prospect status or creating campaigns. This creates a dead end where agents can list prospects by status but cannot move them through the pipeline, making the surface incomplete for full workflow automation.
SEO MCP server for keyword research, live SERPs, backlinks, site audits and AI visibility.
SEO MCP server for keyword research, SERP analysis, audits, and Search Console workflows.
Official Ahrefs MCP server. Access SEO data, including backlinks, keywords, and competitor research.
Official MCP server for Agentwork — delegate tasks to AI agents with human-in-the-loop
