Bill Mongan

Bill Mongan

Professor, Author, Machine Learning Researcher, Educational Consultant, and Scholar of Teaching and Learning; Host: digitalsignature.fm

CSTA Philly

less than 1 minute read

Published:

Drexel is an institutional member of CSTA Philly, an active group in developing curriculum, sharing ideas, and broadening the reach of computing across educational disciplines.

You May Also Enjoy

Giving Every Student Their Own VM: Jetstream2 Classroom Provisioning with OpenStack

11 minute read

Published:

This term, every student in my class has their own Linux VM on Jetstream2, an NSF-funded research cloud available through the ACCESS program. Educational use is free. I use the VMs to run agentic coding tools such as OpenCode inside Docker, so students do not have to install anything on their laptops and I do not pay for cloud machines myself.

A Herdr Docker Devbox: Running Several Coding Agents in One Container with Approval-Based Host Access

99 minute read

Published:

Coding agents are useful because they can read your files and run commands. That also makes them risky to run on a machine with real data on it. A model can misread an instruction, or a file it opens can contain instructions of its own, and an agent running as you can do anything you can do. I run several agents at once, and each one wants access to the same drives. I’d like them to cooperate without all of them being able to see everything.

Herding Agents: Multi-Agent Coordination with Herdr and Per-Agent MCP Credentials

40 minute read

Published:

I now run nine coding agents side by side on a small always-on Ubuntu server: Claude Code, Codex, Antigravity, Grok Build, Hermes, GitHub Copilot, two separate copies of OpenCode, and pi. Herdr keeps them organized, one labelled tab per agent, and brings them back after a reboot. Most of them share one mcpproxy instance for tools. One of those tools talks to a service that needs a personal API key, and only one agent ever holds that key: a deliberately fenced-in OpenCode. The other agents can see the tool, but they can’t authenticate to the service behind it. When they need that service, they hand the task to that OpenCode through Herdr.

mcpproxy: A Config-Driven MCP Host with a Built-In Web UI

5 minute read

Published:

The Model Context Protocol defines a standard way for AI clients to discover and call tools, but standing up a personal MCP server still usually means writing Python glue code, wiring up a framework, and restarting a process every time you add a tool. mcpproxy takes a different approach: every tool provider is a single YAML file, the server reloads tools at startup without any code changes to the host, and a browser-based web UI handles the full provider lifecycle — editing, secret management, and live command streaming — without leaving the browser.