Skip to content

Tutorials

Complete exercises that end in something that works. Follow one start to finish and you will have built a real thing rather than read about one.

Getting started is the one to begin with. It runs entirely in the portal and finishes with a chatbot that still knows your name after you reload the page. Nothing to install — you will need an API key from an LLM provider, and that is the only thing you have to fetch.

  • Getting started — create your first chatbot agent, entirely in the portal, in about twenty minutes.
  • AI Coding: Connecting an MCP host — wire Cursor, Cline, Codex CLI, Copilot, Claude Desktop, Windsurf, or Antigravity to a Hadron app.
  • Chatbot: Designing a multi-stage conversation — design a real three-stage flow (intake → research → summarize) with prompts, extraction specs and routing signals. Ends with a runnable design.
  • Chatbot: End-to-end test — build an agent, design conversations, run chats as test personas, and verify results. Pick Path A (Portal Chat) for production-fidelity, or Path B (Claude Code + MCP) for free iteration.
  • Market-research agent with structured memory — give a memory a typed competitor collection, capture conforming records, run a where + sortProperty query ("series-A competitors funded in the last 90 days, by funding descending"), and audit conformance.