HomeAgentic

Agentic AI

The layer underneath my work. Claude Code sits in the middle, MCP servers connect it to my tools, an Obsidian vault is its memory, and n8n pipelines run on their own.

Pick one to bring it forward, or click a layer.

Agentic Hub. The whole wiring in one place: 9 tools of my own, 10 MCP servers wired in, 8 n8n pipelines on my VPS.

Open the deep dive

Agentic Hub

Claude Code and everything I wired around it, shown live: how a tool call finds its server, every n8n pipeline side by side, and how sub-agents split up the work.

tools in my own MCP server
9
MCP servers wired in
10
n8n pipelines on my VPS
8
Open the Agentic Hub
The hub's hero: Claude Code as a core, with rings for my vault tools, the wired-in MCP servers and the n8n pipelines.
Claude Code
second-brain
  1. callsvault_recallon second-brain
  2. sends{ "task": "why vault_recall is the default search", "limit": 2 }
  3. returnsvault_recall (Hybrid Search) als task-aware Defaultdecision · 1.00
  4. returnsVault Data-Management Iterationdecision · 0.56

A real call to my vault server, titles and scores as they came back.

MCP Servers

A tool becomes a typed function Claude can call: a name, a description, a schema. The deep dive opens both servers I built and the ones I wired in.

built from scratch
2
wired in as clients
10
Open MCP Servers

Second Brain

Plain Markdown, one idea per file, with frontmatter Claude can filter by. It searches the vault by meaning on my own machine before it answers, and writes a new note when something is worth keeping.

A decision note from the vault, rendered: type, project, tags and date above the idea.
One decision note as the deep dive shows it: frontmatter on top, the idea below.
notes in the vault
360+
note types
4
MCP tools to reach them
9
Open Second Brain