For agents

Humans get the pages and developers get the console. AI agents get the whole portfolio as plain text, and a live MCP server they can plug into.

Connect over MCP

Read-only, no sign-in. Add the server to your agent, then ask it something like “What has Shifat built with LangGraph, and how do I reach him?”

Terminal

claude mcp add --transport http shifat https://www.heyshifat.com/mcp

Settings → Connectors → Add custom connector

https://www.heyshifat.com/mcp

~/.cursor/mcp.json

{
  "mcpServers": {
    "shifat": {
      "url": "https://www.heyshifat.com/mcp"
    }
  }
}

.vscode/mcp.json

{
  "servers": {
    "shifat": {
      "type": "http",
      "url": "https://www.heyshifat.com/mcp"
    }
  }
}

get_profileWho Shifat is: bio, current focus, headline numbers, how he builds, stack, and contact details. Start here.

list_projectsEvery project he has built, optionally filtered by category or technology.

get_projectOne project in full, including its case study when there is one.

get_experienceEvery role he has held, with freelance client projects and their reviews.

list_postsHis blog posts, newest first, optionally filtered by tag.

get_postOne blog post in full, as markdown.

get_servicesWhat he offers clients and who each service suits.

get_contactHow to reach him. This server cannot send messages, so pass these details to your user.

searchSearch the whole portfolio (pages, projects, posts, case studies) and get the best matching links.

/llms.txtAn index of every page, in the llms.txt format.

/llms-full.txtThe whole portfolio in one file.

/projects.mdAny page as markdown: add .md to its path (/index.md for home).

Or keep the normal URL and ask for markdown:

curl -H "Accept: text/markdown" https://www.heyshifat.com/projects