Hosted Foglift MCP Server

Local stdio setup for Cursor, Windsurf, and power users

Some clients still expect MCP servers to run as local commands. For those clients, keep using the npm package with an API key. This is the right path for Cursor and Windsurf today, and it remains useful when you want all traffic to originate from your own machine.

1. Generate an API key

Sign up at foglift.io, then go to Dashboard > Developer. API key creation is available on Launch, Growth, and Enterprise plans. Free accounts can use the hosted OAuth MCP flow and public scan_website calls without a key.

2. Add the stdio server

{
  "mcpServers": {
    "foglift": {
      "command": "npx",
      "args": ["-y", "foglift-mcp"],
      "env": { "FOGLIFT_API_KEY": "sk_fog_..." }
    }
  }
}

3. Restart and verify

Restart the client and ask it to call whoami. You should see the authenticated account, plan, default workspace, and token balance in the agent response.

Local npm 1.2.0 tool inventory

The published foglift-mcp 1.2.0 package exposes 15 tools. The package keeps whoami as tool 14 so you can verify account identity before an agent reads usage data or changes monitoring settings. recipe_run is tool 15 and runs packaged workflows.

Current hosted tool inventory

A live tools/list call against https://mcp.foglift.io/api/mcp returned 24 tools on June 11, 2026. The hosted endpoint includes Actions Engine and setup tools that are newer than the original npm-only page described.

Three workflows worth automating

Audit a release candidate

“Use scan_website on the preview URL, then summarize the lowest AI Readiness, SEO, and accessibility signals in a table.”

The agent gets the same structured scan data the dashboard uses, then turns it into a release checklist inside the same session.

Find the next content fix

“Call get_recommendations and get_ai_results. Which prompt has the clearest source and competitor opportunity, and what page should we improve first?”

This pairs the Actions Engine with live AI Visibility history, which is the same dogfooding loop Foglift uses on its own site.

Run a repeatable recipe

“Use recipe_run for the weekly brand health report, then turn the result into a short changelog entry.”

Recipes make recurring AI search checks repeatable. The current tool name is recipe_run, with an underscore, so it works in clients that reject dotted tool names.

The closed-loop pattern: edit, scan, verify in one session

AI search optimization works best as a tight loop. The agent scans a page, identifies the missing citation or readiness gap, edits the page, re-scans after deploy, then pins the prompt that the page should win.

Foglift uses this loop internally. A June 12, 2026 AI Visibility baseline showed that broad discovery answers repeatedly cite tool roundups and developer workflows. Public API, CLI, and MCP documentation gives engines precise capability evidence to extract while giving builders a working path from monitoring data to implementation.

Foglift MCP vs. other AI search optimization tools

MCP support across the AI search optimization category from the May 2026 parity review:

See the supporting category review in Best AI Search Tools with MCP Integration 2026.

Frequently asked questions

Sources

• Foglift live MCP endpoint. https://mcp.foglift.io/api/mcp, https://mcp.foglift.io/.well-known/oauth-authorization-server, and https://mcp.foglift.io/.well-known/oauth-protected-resource. Re-verified June 11, 2026.
• Anthropic. Introducing the Model Context Protocol (November 2024).
• Anthropic. modelcontextprotocol/servers. Reference servers and client list.
• Foglift. foglift-mcp on npm. Local stdio package and version history.