Docs

Enpitech MCP App

The Enpitech MCP App brings our tools into Claude, ChatGPT and any other MCP client. AgentReady scans a public website and scores 0-100 how ready it is for AI agents. The AI Hub search finds the skills, MCP servers and workflows we use for frontend work. The newsletter and contact form let you subscribe or reach the Enpitech team without leaving the conversation. The scanner and the forms open as interactive views in hosts that support MCP Apps, and work as plain tools everywhere else.

Connect

Claude (web, desktop and mobile)

Open Settings → Connectors. Find Enpitech in the directory and connect it, or choose Add custom connector, name it Enpitech and paste the server URL. Then turn it on for a chat from the tools menu.

Claude Code

claude mcp add --transport http enpitech https://mcp.enpitech.dev/mcp-app

ChatGPT

Find Enpitech in the ChatGPT app directory and connect it. On plans with developer mode, you can also add it yourself as an app with the server URL and no authentication.

VS Code, Cursor and other MCP clients

Add a remote HTTP server with the server URL. For VS Code, in .vscode/mcp.json:

{
  "servers": {
    "enpitech": { "type": "http", "url": "https://mcp.enpitech.dev/mcp-app" }
  }
}

For Gemini CLI, in settings.json:

{
  "mcpServers": {
    "enpitech": { "httpUrl": "https://mcp.enpitech.dev/mcp-app" }
  }
}

Clients without MCP Apps support call the same tools and show the results as text instead of the interactive views.

Try it

• "How ready is stripe.com for AI agents?"
• "Compare how agent-ready vercel.com and netlify.com are."
• "What should I fix first to make my site readable to AI?"
• "Find me a Claude Code skill for turning Figma designs into React."
• "What MCP servers does the AI Hub recommend?"
• "Subscribe me to the Enpitech newsletter."
• "I want to talk to Enpitech about getting our backend engineers and PMs shipping frontend."
• "I'd like private Claude Code training for my frontend team."

Tools

• agent_readyRead-only

  Opens the AgentReady scanner as an interactive view. It runs 21 checks on the site you name and shows a 0-100 score, a breakdown across five categories, and fixes ranked by the points each one recovers.

• agentready_scanRead-only, fetches public pages

  Runs the same scan and returns the result as data, with no view. Used when you compare several sites or want the raw results.

• email_agentready_reportWrite: sends an email, saves an enquiry

  Emails you the full report from the results view and saves your request as an enquiry to Enpitech. Only the view calls it, after you fill in the form. The assistant cannot call it on its own.

• ai_hub_searchRead-only

  Searches the Enpitech AI Hub, our curated catalog of Claude Code skills, MCP servers and AI tools, and agentic workflows for frontend teams. Returns each match with a short take, an overview, the install command when there is one, and a link to its page.

• newsletterRead-only

  Opens a signup form for the Enpitech newsletter (frontend, AI and how teams ship) as a view. You type your own email there.

• subscribe_newsletterWrite: adds you to the mailing list

  Subscribes the email you typed into the newsletter form. Only the view calls it, after you submit. The assistant cannot call it on its own.

• contact_formRead-only

  Opens Enpitech's contact form as a view in the chat. It sends nothing by itself.

• submit_contactWrite: saves an enquiry

  Sends the enquiry you submit (name, email, message, optional phone and company) to the Enpitech team.

Every tool declares its behavior with MCP annotations (read-only, destructive, open-world), so your client can ask before anything is sent.

What AgentReady checks

21 checks across five weighted categories: Discovery & Access, Readable for Agents, Structured Data, Actions & MCP, and Trust & Safety. It covers things like robots.txt rules for AI crawlers, llms.txt, schema.org markup, and whether the site exposes MCP or other agent actions. You can also run it in the browser at enpitech.dev/agent-ready.

Data and privacy

• The server receives only the inputs of the tool that is called. It never sees the rest of your conversation, your chat history or your AI account.
• Scans fetch only public pages of the site you name and change nothing on it. The address is not stored.
• AI Hub searches run over our published catalog. The search terms are not stored.
• Newsletter signups add the email you typed to our mailing list in Mailchimp, tagged as coming from the MCP App. Every email has an unsubscribe link.
• Enquiries and report requests store the details you typed into the form as an enquiry to Enpitech, tagged with the form it came from (and, for a report, the site and its score). We use them only to answer you.
• To limit abuse, we use the IP address the request comes from (the AI platform's, not yours) and, on ChatGPT, the anonymized user id ChatGPT sends for rate limiting. Both are held in memory for a few minutes and are not saved.
• The MCP App sets no cookies and needs no login.

Full details are in our privacy policy. Use of the service is subject to our terms.

Limits and troubleshooting

• Rate limits. Scans, report requests and newsletter signups are rate limited to protect the scanned sites and our systems. Per minute: 10 scans per ChatGPT user, 60 per AI-platform address (Claude sends no user id, so its users share this one), 10 scans of any one site, and 120 in total. Report requests and signups have smaller budgets. If a scan returns ratelimited, wait the seconds it names and try again.
• Bot protection. Some sites block automated requests. The result is then marked limited: the per-check breakdown still helps, but the overall score is not meaningful.
• Unreachable or slow sites return unreachable or timeout. Check the address and that the site is public.
• Opening the server URL in a browser shows "Method Not Allowed". That is expected: MCP clients talk to it with POST requests.

Support

Questions, bugs or security reports: email [email protected]. The MCP App is built and run by Enpitech, which accelerates frontend delivery for product teams, including MCP apps and agent-ready interfaces for their own products.