Docs · MCP

Actuity in your coding tool.

actuity mcp runs an MCP server, so Claude Code, Cursor, VS Code, Windsurf and Claude Desktop can scan your site, read the fix prompts, check a fix after you deploy it and run agent analyses, without leaving the editor.

You need an API key for everything except scanning. Using claude.ai or ChatGPT instead? See claude.ai and ChatGPT: no key or install.

Claude Code

Terminal

claude mcp add actuity -e ACTUITY_API_KEY=act_... -- npx -y actuity mcp

Cursor, Windsurf and Claude Desktop

Add this to the tool’s MCP settings file: .cursor/mcp.json in your project for Cursor, or the tool’s own settings for Windsurf and Claude Desktop.

mcp.json

{
  "mcpServers": {
    "actuity": {
      "command": "npx",
      "args": ["-y", "actuity", "mcp"],
      "env": { "ACTUITY_API_KEY": "act_..." }
    }
  }
}

VS Code

VS Code can ask for the key instead of storing it in the file. Add this to .vscode/mcp.json:

.vscode/mcp.json

{
  "inputs": [
    { "type": "promptString", "id": "actuity-key", "description": "Actuity API key", "password": true }
  ],
  "servers": {
    "actuity": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "actuity", "mcp"],
      "env": { "ACTUITY_API_KEY": "${input:actuity-key}" }
    }
  }
}

Don’t commit a key in a project file. Use your tool’s own secret storage where it has one.

Tools

Tool	What it does
scan_site	Scans a site and lists the problems, with a scan id.
get_scan	A scan’s result by id, or the full report as Markdown.
get_fixes	Fix prompts: a site’s tracked fix list with fix ids, or one scan’s problems.
recheck_fixes	After a deploy, checks fixes again and says which are fixed. Paid plans.
list_sites	The sites in your workspace.
suggest_agent_test	Turns a task, like “Find the Pro plan price”, into ways to test it. Uses no credit.
run_agent_test	Runs one: our read-only agents try the task. Uses one agent analysis.
get_agent_test	An agent analysis result by its run id.

Without a key, scan_site, get_scan and get_fixes with a scan id work like the free scan; the rest need one. Scans and analyses can take a few minutes: a tool waits up to 240 seconds, then hands back an id to check later. actuity mcp --timeout <sec> changes the wait.

Before an agent analysis can run from a tool, approve the site once on the website: open it in your dashboard and confirm you own it or may test it. A key can’t give that approval.

Try it

Once it’s added, ask your coding tool something like:

“Scan example.com with Actuity and fix the top problem.”
“I deployed the fixes. Re-check them with Actuity.”
“Run an Actuity agent analysis: find the price of the Pro plan on example.com.”

Next: claude.ai and ChatGPT →