SEOBolt exposes audit, issues, fix-plan, and checkout tools over MCP. Your agent can inspect the crawl, update your repo, and trigger another crawl to verify the result.
Crawler-backed checks, not marketing copy: audits now emit AI Search issues into dashboard reports, fix plans, REST API responses, and MCP tools.
Create an API key and configure the SEOBolt MCP server.
Run seo_start_crawl from the agent session.
Call seo_fix_plan and let the coding agent patch the repo.
Run seo_audit after deployment to confirm the health score moved.
Traditional SEO still matters, but AI search adds new failure modes: blocked AI crawlers, missing entity context, thin llms.txt files, and pages that answer engines cannot cite. SEOBolt reports those as first-class audit issues with fix templates.
# Remote server: nothing to install (Claude Code)
claude mcp add --transport http seobolt https://seobolt.app/mcp \
--header "Authorization: Bearer sk_live_YOUR_KEY"
# Cursor: ~/.cursor/mcp.json
{
"mcpServers": {
"seobolt": {
"url": "https://seobolt.app/mcp",
"headers": { "Authorization": "Bearer sk_live_YOUR_KEY" }
}
}
}
# Claude Desktop (or any client): run the npx package locally (stdio)
{
"mcpServers": {
"seobolt": {
"command": "npx",
"args": ["-y", "mcp-server-seobolt"],
"env": {
"SEOBOLT_API_KEY": "sk_live_your_key"
}
}
}
}Every MCP tool maps to an authenticated API endpoint with API-key rate limits.
The agent does not need to interpret a PDF; it receives snippets and affected URLs.
SEOBolt is built around re-crawling after fixes, not stopping at the audit.
To run an SEO audit over MCP from Claude Code, add the remote server shown above: claude mcp add --transport http seobolt https://seobolt.app/mcp with your API key in the Authorization header. Nothing is installed locally. Clients that only launch local commands, like Claude Desktop, can use the mcpServers block instead: the command is npx -y mcp-server-seobolt and your API key goes in SEOBOLT_API_KEY. Agents that have no browser can get a key from POST /api/agent-signup.
Once connected, ask Claude to call seo_start_crawl for your project, then seo_audit for the health score and seo_fix_plan for every issue with its code snippet and affected URLs. The agent edits your repository, you deploy, and it starts another crawl to confirm the issues are gone.
SEOBolt's Model Context Protocol (MCP) server enables AI coding agents (Claude, Cursor, Codex) to run site audits, retrieve structured fix plans with code snippets, patch repository files, and trigger re-crawls directly inside your coding workflow.
The MCP server has 22 tools, including seo_projects, seo_start_crawl, seo_audit, seo_issues, seo_fix_plan, seo_fix_suggestions, seo_content_score, and the Google Search Console tools. The remote endpoint and the npx package expose the same list.
Send your SEOBolt API key as an Authorization: Bearer header to the remote endpoint at https://seobolt.app/mcp, or pass it as SEOBOLT_API_KEY in the environment of the npx mcp-server-seobolt package.
Run claude mcp add --transport http seobolt https://seobolt.app/mcp --header "Authorization: Bearer sk_live_YOUR_KEY" in Claude Code, or add a seobolt entry to Claude Desktop's mcpServers config that runs npx -y mcp-server-seobolt with SEOBOLT_API_KEY set to your key. Then ask Claude to start a crawl and fetch the fix plan.
Not yet. The remote endpoint authenticates with a Bearer API key and does not offer OAuth sign-in, which one-click connectors in claude.ai, Claude Desktop and ChatGPT require. Use Claude Code, Cursor or Codex with the key in a header, or run the npx package from Claude Desktop's config.
Check your homepage for AI search access, llms.txt quality, crawler rules, structured data, Open Graph, canonical tags, and the first fixes to apply.
Run the audit