MCP server

Talk to my content from your own AI.

A public, read-only MCP server with my articles, architecture patterns, glossary and the cost calculator. Connect it to Claude or another compatible tool and ask about AI architecture with my sources.
endpoint /api/mcp
transport streamable http · stateless
access read-only · 6 tools
Claude (claude.ai / desktop)
  1. 1Open Settings → Connectors
  2. 2Click "Add custom connector"
  3. 3Paste the URL below and save
URL
https://wictorvargas.com/api/mcp
Claude Code
terminal
claude mcp add --transport http wictor-vargas https://wictorvargas.com/api/mcp
Other MCP clients
mcp.json
{
  "mcpServers": {
    "wictor-vargas": {
      "type": "http",
      "url": "https://wictorvargas.com/api/mcp"
    }
  }
}
Available tools
  • search_articles

    Search Wictor Vargas's articles on AI and systems architecture (RAG, MCP, agents, governance, cost, observability). Returns the most relevant passages with links.

  • get_article

    Get the full text (Markdown) of one article by its slug, as returned by search_articles or list_articles.

  • list_articles

    List all published articles, newest first, with date, summary and link.

  • get_pattern

    Get an AI architecture pattern (problem, solution, when to use, trade-offs). Omit slug to list all patterns.

  • glossary

    Look up AI architecture terms (MCP, RAG, guardrails, eval, trace...). Omit term to get the whole glossary.

  • estimate_ai_cost

    Estimate the monthly API cost of an LLM workload from request volume and token sizes, using reference prices.

Try asking
  • “According to Wictor Vargas, how do I limit what an agent can do in production?”
  • “How much would 5,000 tasks a day with 8,000 input tokens cost on Sonnet?”
  • “Explain the MCP gateway pattern and when to use it.”

The server requires no login and does not log the content of calls.