Jeremy Neely Text Jeremy

AI & Agents · Beta

Secure AI Advanced Chat

Saved-prompt workspaces for Claude through the Hatz AI gateway.

Built for: AI and solution engineers at The AME Group

A saved App drafting a statement of work, with the activity log (sample data).
A saved App drafting a statement of work, with the activity log (sample data).

The problem

AME engineers build structured prompt packages, such as a Statement of Work generator driven by YAML instruction files. The standard chat window made them awkward to reuse, and nobody could see what a long conversation cost until it was over.

What I built

I built a power-user Claude client that saves each prompt package as a named App with its reference files. It streams replies through the Hatz AI gateway, tracks tokens live and encrypts its settings. It exists as both a desktop app and a browser edition.

How it works

  1. Enter the Hatz key and gateway once; settings are encrypted.
  2. Create an App with instructions and reference files.
  3. Send; files become Claude content and the reply streams in.
  4. Watch each API step in the activity log.
  5. Track chat and session tokens against a warning level.
How Secure AI Advanced Chat is wiredSettings are encrypted with a password. The same prompt package runs in a desktop and a browser edition.
Comes inInside Secure AI Advanced ChatGoes out Saved App instructions plus files Hatz key & gateway entered once, encrypted Build request files become content Stream reply through the Hatz gateway Track tokens chat and session Streamed answer tabs or side by side Activity log every API step Token warning at a set level

Built with

  • Python 3.10+
  • customtkinter
  • Anthropic Python SDK
  • httpx
  • cryptography
  • Flask
Text Jeremy about this Next: Hatz Admin Toolkit All 38 tools Texting opens your Messages app with a note already written. He replies himself.

More screens

The web edition drafting a statement of work from a request (sample data).
The web edition drafting a statement of work from a request (sample data).
The agents list, each with its own instructions and files (sample data).
The agents list, each with its own instructions and files (sample data).
Configuration for model, output limit and token warning level.
Configuration for model, output limit and token warning level.
The web edition's start screen before an agent is loaded.
The web edition's start screen before an agent is loaded.

More of Jeremy's work

All 38 tools, by category