ToolsNest MCP Server
ToolsNest also speaks MCP. Point a compatible AI agent at our server and it can run an SEO audit, check a robots.txt file, or score readability using the same code that powers the tools on this site, no scraping our pages or guessing at our output format.
What this is
The Model Context Protocol is an open standard that lets AI agents call external tools through a common interface, instead of every provider inventing its own plugin format. ToolsNest runs a remote MCP server at https://toolsnest.io/mcp over Streamable HTTP, the transport the current MCP spec recommends for a hosted server like ours.
Nine tools are exposed today, all of them backed by the exact same functions our website calls, not a copy, not a simplified version. If we improve the SEO audit logic for the website, the MCP tool improves with it.
Available tools
toolsnest_seo_auditFull on-page SEO audit of a public URL, with a 0 to 100 score.
toolsnest_compare_seoAudits two URLs and compares their scores side by side.
toolsnest_check_robotsFetches and checks a domain's live robots.txt for problems.
toolsnest_analyze_keyword_densityTop keywords by frequency and density in a block of text.
toolsnest_check_readabilityFlesch reading ease and grade level for a block of text.
toolsnest_count_wordsWord, character, sentence, and reading time counts.
toolsnest_count_charactersCharacter counts against Twitter, LinkedIn, and Google Ads limits.
toolsnest_generate_meta_tagsTitle, description, Open Graph, and Twitter Card HTML.
toolsnest_generate_robots_txtBuilds a robots.txt file from user agent rules and sitemaps.
The Website Speed Checker and Plagiarism Checker are not exposed over MCP. Both call metered third party APIs, and we do not want an agent able to spend that budget on our behalf.
How to connect
Add the server to your client's MCP configuration with the endpoint URL and a bearer token in the request headers:
{
"mcpServers": {
"toolsnest": {
"url": "https://toolsnest.io/mcp",
"headers": {
"Authorization": "Bearer YOUR_TOOLSNEST_MCP_TOKEN"
}
}
}
}That works for Claude Code, Claude Desktop, and any other client that accepts a Streamable HTTP server with custom headers. If your client only supports the MCP Inspector or an SDK client, the same URL and header work there too.
Getting a token
The server is not open to anonymous traffic. Every request needs a bearer token, and we hand those out manually for now rather than running self serve signup.
How to request accessRate limits
Requests are limited per token so one client cannot exhaust the service for everyone else. If you hit the limit, the response tells you how long to wait before trying again.
How we keep it safe
- Fetch toolsTools that fetch a URL, the SEO audit, the comparison tool, and the robots.txt checker, refuse to request localhost, private networks, and cloud metadata addresses, and revalidate every redirect before following it.
- Auth todayAccess is a shared bearer token, checked in constant time, not OAuth. If you were told otherwise, that is a mistake, and we would rather say so plainly than let a client assume a compliance level we do not have yet.
- No writesEvery tool reads and analyzes. None of them write to our database, run a shell command, or fetch and hand back an arbitrary page verbatim.