AgenticSystemCore Demo Node
SummaryA live demonstration node of the AgenticSystemCore reference engine — the second node of the Agentic Knowledge Web — running on a small sample of well-known agent-system patterns.
This site is software at work. It is a live demonstration node of the AgenticSystemCore reference engine, the open-source program that turns a folder of Markdown files into a knowledge node that people and agents can find, check and cite. It is the second node of the Agentic Knowledge Web, beside agenticsystemcore.com: each of the two names the other in its discovery file. Every page and every file here was built by the engine as static files: no application server and no database run behind it.
What the software does on this node
- Discovery. One file tells an agent what this node is and where everything else is, with a digest for every file it names:
/.well-known/knowledge-linkset. - A typed graph. The items and the typed links between them, in the format's published vocabulary: JSON-LD, Turtle and N-Quads.
- Text for language models.
llms.txt,llms-full.txtand the chunk exportchunks.jsonl, every chunk with its source address. - Search. The search page runs in your browser over a static index; nothing leaves this site.
- Tools for assistants in the browser. Every item page registers seven page tools that an assistant in the browser can call to search, read, follow links, compose and cite (how they work).
- Machine views. Beside every page sit a Markdown and a JSON-LD copy — for example the Markdown and JSON-LD of Blackboard.
- Provenance and checked references. Each item says where it came from and who answers for it, and lists the public sources it rests on, each checked against the source.
- A change ledger. Every published change is a line of
ledger.jsonl; the changelog lists the tagged versions. - Composition. The compose page turns the items you tick into a starting set of seven harness files, computed in your browser. The skill packs give one file per cluster for agent tool folders.
The sample content
The content is a small sample of well-known agent-system patterns, each described from public sources and grouped into clusters. It is here so the software has real material to work on: every item carries its references, checked against those sources, and a diagram whose text source travels beside it. Browse the concepts or the clusters, or start from the search page.
Run your own node
The engine is the npm package agentic-system-core. A Python package of the same name on PyPI checks and reads published nodes. The source is in the engine repository. The published items of this node — the Markdown files, the diagram sources and the configuration — are in the agsc-demo-node repository. To begin, read start here and publish a Level-0 node on agenticsystemcore.com.
What comes next
Planned next: a stated meaning for every typed link, so that the links become a semantic layer software can check, and a graph view of the links — see the roadmap.