Skip to main content

Agentic AI Overview

neonto ships an in-process Model Context Protocol (MCP) server that exposes its RDF tooling to AI coding agents, using the same engines that power the editor.

Why it exists​

An AI agent working on Turtle, SHACL, SPARQL, or YARRRML files benefits from the same validation and query engines a developer uses inside the editor, rather than guessing at RDF syntax or vocabulary terms from training data alone. The neonto MCP server exposes that tooling — SHACL validation, SPARQL execution, vocabulary lookup, YARRRML/RML processing, and live diagnostics — as MCP tools an agent can call directly.

Requirements​

  • Desktop VS Code only. The MCP server runs in the Node extension host and is not available in VS Code for the Web.
  • License and login required. Like all neonto features, the server is entitlement-gated: requests from an unauthenticated session are rejected with an HTTP 401.

How it runs​

The server is an in-process Streamable HTTP MCP server, started automatically when the extension activates. It listens on http://127.0.0.1:<port>/mcp on localhost only.

  • The port is controlled by neonto.mcp.port (default 6789, range 1024–65535).
  • If the configured port is already in use, the extension automatically binds a free port instead, so the server always starts.
  • The bound URL is written to a temporary file on activation and can also be copied to the clipboard, so external agents can discover it without you having to inspect logs. See Connecting an AI Agent.

Capabilities​

A connected agent can:

  • validate and repair RDF data against your SHACL shapes
  • model ontologies using real vocabularies from Linked Open Vocabularies
  • build data pipelines and run ingestion (YARRRML/RML ETL)
  • query your graph and configured SPARQL endpoints
  • pull up concise syntax guides on demand

See Agentic Features for each capability with example prompts.