AI agents, MCP & developer experience

How to use this site with LLMs, Model Context Protocol, Cursor skills, and machine-readable context.

Built for humans and AI assistants

This page summarizes how recruiters, developers, and automated agents can reliably use this portfolio: search, structured JSON, MCP-oriented discovery, and Cursor-style agent skills.

Quick reference

Resource	Purpose
Site search	Press ⌘K / Ctrl+K — GitBook-style palette with context preview on the selected row and Ask AI to pass that snippet into the chat.
Ask AI chat	Floating message button → slide-over chat. Uses `POST /api/ask-ai/` when routes exist; otherwise offline FAQ from the same knowledge as this site.
skills.md	`/skills.md` — Cursor / MCP / agent summary of skills, employers, and domains (pair with `llms.txt`).
LLM context JSON	`GET /api/llm-context/` — concise facts, FAQs, and tone guidance for models (when the deployment exposes API routes).
llms.txt	`/llms.txt` — lightweight index of key URLs for crawlers and tools that follow the convention.
Agent resources	`/.well-known/agent-resources.json` — links + `mcp_context_bundle` for bulk fetch.
Markdoc ↔ MDX	`/docs/developers/markdoc-and-mdx/` — how Markdoc-style tags map to MDX components here.
Skills (portfolio)	`/docs/skills/` and `/skills/` — human-facing skill matrix and narrative.
Cursor agent skills	Optional `.cursor/skills/` on your machine (see below); this site documents the pattern, not private skill files.

Static hosting note. This project uses Next.js static export for production. /api/* routes are not present in the pure static out/ bundle unless your host adds serverless functions. Use /llms.txt, agent-resources.json, on-page structured data, and /search/ when APIs are unavailable.

Model Context Protocol (MCP)

MCP connects assistants to tools and data sources. This site does not ship a hosted MCP server, but it supports MCP-friendly workflows:

HTTP context — Point your MCP tool or custom server at /api/llm-context/ (when available) or mirror /llms.txt into your server’s resources.
Structured excerpts — MDX pages include hidden application/json+ai-context blocks (via the AIContext component) for entity-level hints.
Discovery — /.well-known/agent-resources.json lists canonical URLs so an agent can fetch documentation without guessing paths.

Checklist: wiring an MCP server

0 / 4 completed

Fetch `/llms.txt` or `agent-resources.json` on a schedule or on first use.

Prefer `/api/llm-context/` for FAQ-style answers when the route exists on your deployment.

Fall back to fetching specific `/docs/...` pages for deep technical detail.

Cursor skills & DX

Agent skills (e.g. under .cursor/skills/) package repeatable instructions for your editor’s AI. This repository may include skills for Freshworks FDK, Next.js patterns, etc.—those live in your clone, not on the public HTML.

Developer experience on this site:

Cmd/Ctrl+K global search across portfolio + docs, with GitBook-style context preview and Ask AI handoff.
Ask AI slide-over (FAB) with optional search/MCP context attached to each message.
Trailing-slash URLs and stable static routes for predictable automation.
MDX docs with Markdoc-class patterns — see Markdoc vs MDX.
OpenAPI sections under /docs/api/ for interactive API exploration.
RSS at /rss.xml for feed-aware tools.

Asking an AI about this portfolio

When prompting a model, you can say:

Use the JSON from {origin}/api/llm-context/ if available; otherwise read {origin}/llms.txt and the /docs/developers/ai-mcp page. Cite only facts present there.

For adding context to a thread without an API:

Open Search (⌘K / Ctrl+K) and find the right doc.
Copy the page URL or export text from the doc.
Paste into your assistant with an instruction to stay within that scope.

Documentation