JMN AI — Operator's Manual
Build an agent. Read it like a manual, not a marketing page.
This is the working manual for JMN AI — what an Agent is, how to build one column by column, every setting you can change, and the advanced behavior (collections, sync) worth understanding before you rely on it. Start at 01 if you're new; jump straight to Reference if you just need a setting explained.
About this draft
This is a local working copy. Screenshots are placeholders until captured — each one shows the exact filename it's waiting on, right where it belongs.
01 — Getting Started
1.1
What is an Agent?
The core model: Agent → Properties → Rows, and the spreadsheet metaphor.
1.2
Creating Your First Agent
Step by step: new agent, first file, first prompt, first result.
02 — Building Agents
2.1
Properties Explained
Input vs. output, and every output type.
2.2
Tools & How They Work
RAG retrieval, full-document mode, and the Python tool.
2.3
Prompting 101
@PropertyName, @output references, and @class labels.
2.4
Model Selection Guide
Which model to use, and when it matters.
2.5
Settings Reference
Every property setting, what it does, and its default.
03 — Advanced
3.1
Collections (Sub-Agents)
What happens when one column gets many files.
3.2
Keeping Results in Sync
Why edits never silently cascade, and what "stale" means.
3.3
Tagging & Classification
Tag columns, options, and manual overrides.
04 — Account & Team
4.1
Inviting Teammates
Seats, roles, and how invites work.
4.2
Usage & Credits
Reading your credit balance and what happens at zero.