A unified, microservices-based command center that replaces the patchwork of PSA, RMM, XDR security, multi-provider AI assistance, voice/SMS, and billing tools — running on your own infrastructure with zero per-technician SaaS seat fees.
Every screen below is captured directly from the live Glass application running against our sanitized demo tenant — no mockups or AI-generated concept art. Click any tab to switch screens, or click a screenshot to inspect it in full 1080p resolution.
Glass treats documentation as an engineering artifact. Every module has both a source-verified written User Guide and an automated suite of task-based walkthroughs and AI-narrated video tutorials.
Bite-sized, task-specific video walkthroughs captured from the live platform by the Playwright tutorial builder, complete with natural AI voiceover, synchronized WebVTT closed captions, and numbered step-by-step procedures.
Complete reference documentation covering prerequisites, interface tables, everyday tasks, bulk operations,
permissions, and exact error messages. Continuously checked against the PHP/Go/Python source code by user_guide_verify.php.
Click any screenshot card below to open the full 1920×1080 capture in the lightbox viewer.
Five role views where every metric opens its exact filtered record list, plus Kopia backup status and unbilled time.
High-density ticket queue with instant server-side search, inline status/priority dropdowns, and Kanban board toggle.
Scoped Q&A across client ticket history, Dual-LLM sandbox verification, and on-demand provider/model selection.
Private operator sessions with live MCP tool access to query tickets, assets, warranties, and alerts in real time.
Configure primary DGX Spark (`ollama`), secondary workstation GPU (`ollama_local`), and cloud failover (`Gemini`, `Claude`, `OpenAI`).
Full asset inventory with live online status and instant Remote Desktop, Terminal, and File Browser launch buttons.
Correlates SentinelEDR, Huntress, and Defender telemetry with MITRE ATT&CK mapping and one-click host isolation.
Monitors support mailboxes, classifies urgency, threads replies onto existing tickets, and creates new tickets in one click.
Slide-over customer support chat drawer connected to endpoint tray agents, alongside Go WebSocket internal team chat.
Operators can search and read all 37 verified User Guides directly inside Glass without leaving their workflow.
Traditional MSP stacks stitch together a cloud PSA, a separate RMM, a bolt-on ticketing inbox, a third-party documentation wiki, an external billing sync, and per-seat AI add-ons — paying per technician per month for each one, while critical operational data lives on someone else's servers.
Glass was engineered from the ground up at Computer Resources Unlimited (CRU) as a single, cohesive, self-hosted platform. Every ticket, endpoint, PowerShell remediation, Wazuh/Sentinel security alert, UniFi call, Kopia backup, and invoice lives in one relational model — enriched by Silica (AI) with multi-provider routing across local NVIDIA DGX Spark hardware, secondary workstation GPUs, and frontier cloud APIs.
Run 100% air-gapped on an NVIDIA DGX Spark (128GB unified memory) and local workstation GPUs, with a Dual-LLM security sandbox that scans untrusted emails and logs before tool invocation.
Assign lightweight local GPUs (`ollama_local`) to fast triage roles, heavy reasoning to DGX Spark (`ollama`) or frontier models, and define an automatic failover chain (`AI_BACKUP_CHAIN`).
PHP 8.3 (strict types) for core workflows, Go for high-concurrency telemetry & WebSockets, Python/FastAPI for AI orchestration & benchmarking, and Rust for edge agents.
Sixteen+ purpose-built services orchestrated with Docker Compose, sharing a unified PostgreSQL + pgvector data layer, Redis event bus, and multi-provider Silica (AI) gateway.
Explore the interactive Tutorials and User Guide portals right now, or reach out to schedule a live technical walkthrough.