MCP

MCP server

SEO AI Regent exposes a Model Context Protocol server so AI agents — Claude Code, Claude Desktop, Cursor, or anything MCP-capable — can query your SEO data and run audits and scoring on your behalf. The server is stateless over streamable HTTP and authenticated with an API key. Every tool is scoped to the workspaces you own, and plan entitlements apply exactly as they do in the dashboard.

Create a key at Account → API Keys, then connect:

# Claude Code
claude mcp add --transport http seo-ai-regent \
  https://seo-ai-regent.com/api/mcp \
  --header "Authorization: Bearer sar_your_key_here"

# Any MCP client (streamable HTTP, stateless)
POST https://seo-ai-regent.com/api/mcp
Authorization: Bearer sar_your_key_here
Content-Type: application/json

Which MCP tools are available?

The server exposes 9 tools: they list Workspaces and Sites, read Site Health Scores, audit issues, Search Console opportunities, rank trends and Citation Checks, run audits, and score drafts live.

list_workspaces

Your Workspaces with site and keyword counts. Start here.

list_sites

Sites in a Workspace with verification status and latest-audit summary.

get_site_health

Site Health Score (0-100) with per-severity deductions and issue-type counts.

get_audit_issues

Technical Seo Issues with remediation guidance, filterable by severity and state.

run_audit

Run a sitemap-fed Technical SEO audit (up to 50 pages; plan-gated).

get_opportunities

Search Console opportunities: striking distance, CTR gap, decay, cannibalization.

get_rank_trends

Recent Rank Tracking positions for a Workspace's Keywords.

get_citations

Recent AI-search Citation Checks. ChatGPT, Perplexity, Claude and Gemini are each checked through an API model with web search, not in the consumer apps.

score_content

Score draft content live: Content Score and GEO Score with signal breakdowns.

Scheduled rank checks and citation checks are paused, so get_rank_trends and get_citations return only results already recorded, which are not updating.

Scores returned by the tools come from the scoring engine the dashboard uses, with one difference: score_content counts only site-relative paths (such as /pricing) as Internal Links, while the editor also counts full URLs to your Sites. Publishing is blocked unless the Content Score reaches 70/100, the GEO Score reaches 70/100, and no signal is critical (below 40); either score below 50/100 is a hard block. Every Site Health Score comes with its per-severity deduction breakdown.

Agent Skills

Ready-made skills that teach an agent multi-tool SEO workflows on top of the MCP tools. Download one and drop it into your agent's skills directory (for Claude Code: .claude/skills/<name>/SKILL.md).

seo-health-check

Audit a site, read its Site Health Score, and produce a prioritized fix plan.

content-optimizer

Iterate a draft against score_content until it clears the 70-point Review Gate.

opportunity-triage

Turn opportunities, rank trends, and citation checks into this week's top-5 action list.