AI agents work alongside our engineers on every part of the platform: planning, writing code, testing and documenting. Engineers stay in charge, and every change is reviewed and tested. The result is projects delivered faster, more consistently and better documented.
Our AI agents (Claude Code, Codex, Copilot) don't guess. They read the repo's own instructions first: the architecture, the coding rules and the docs that must change. Then they work through the same steps an engineer would, much faster.
Agents can work in parallel on separate copies of the codebase, so several features progress at once.
AI speeds up the engineers; it doesn't replace them. Four guardrails keep the quality up.
Engineers set direction, review every pull request, and approve every release.
800+ API tests, integration tests against a throwaway database, and 125 UI spec suites.
Security and quality rules live in the repo, so AI tools follow them every time.
44 docs pages are updated with every feature, so knowledge never goes stale.
Examples of the rules every AI agent reads before it touches the code:
Commits to the platform each month. The peaks are major feature pushes, each one tested and released through the same pipeline.
The hard, risky, repetitive parts are built once and tested on every project, so engineering time goes into what's unique to each client.
Mark a method with [ApiTargetMethod] and the framework discovers it and exposes it to the touch panel automatically.
For the client: Panel features take minutes to wire, not hours of join numbers.
proAV.Net.Core/UI/ApiTargetMethodAttribute.csA .NET 8 library on every Crestron 4-series processor — web server, real-time hubs, cloud connector, config, monitoring and scheduling built in.
For the client: Every project starts with connectivity, monitoring, cloud, config and UI plumbing already built and proven.
proAV.Net.Core/Base classes and drivers for displays, cameras, audio, TCP/serial, Shelly, Q-SYS discovery, BACnet, MQTT, sensors and PDUs.
For the client: Integration work is reused across projects, not rewritten for each one.
proAV.Net.Core/DeviceSupport/Turns an iPad into a room panel, with dedicated room-lock mode, PIN exit and unattended recovery.
For the client: A lower-cost panel option that still behaves like an installed touch panel.
docs/docs/panel-app/dedicated-panel-mode.mdAn Angular library for Crestron CH5 panels — live connection services, guards, theming and ready-made components.
For the client: Modern, consistent, brandable panel UIs on every project.
user-interface/projects/proav-ui-lib/Mark a method on the processor with [ApiTargetMethod]. The framework discovers it and the touch panel can call it straight away, with no join numbers or signal mapping. This is real code from our reference app.
[ApiTargetMethod("GetRooms")]
public object GetRooms()
{
var rooms = UxEnvironment.GetRooms();
return rooms.Select(room => new
{ room.Id, room.Name, room.ScreenName });
}this.api.invoke('GetRooms').subscribe(rooms => ...);Sooner. Shared building blocks and AI-assisted engineering shorten delivery.
Consistent. Every room runs the same tested platform, so behaviour and support are predictable.
Documented. Documentation is part of every change, not an afterthought at handover.