The AEM AI Assistant Adobe didn't build. Built by XSC. For XSC.
A page-aware AI assistant for AEM Edge Delivery Services — grounded in product truth, powered by the same skills and agents that ship with aemcoder.
Project Compass is a browser-based AI assistant that understands AEM Edge Delivery Services at a deep level. It reads the page you're looking at, knows the block library, speaks the XSC playbook, and helps pre-sales engineers demonstrate AEM capabilities with authority.
| Layer | Source | Size |
|---|---|---|
| AEM System Prompt | EDS architecture, patterns, terminology | ~18 KB |
| AEM Skill Knowledge | Distilled from 24 aemcoder skills (72 files) | ~8 KB |
| XSC Playbook | Revenue motions, objection handling, demo patterns | ~7 KB |
| Customer Context | Dynamic — reads the current page in real time | Variable |
- Open Project Compass
- Enter your Claude API key in Settings
- Navigate to any AEM EDS page
- Ask Compass anything — it sees what you see
- Runtime: Pure browser — no server, no build step
- AI: Claude (Anthropic) via direct API
- Content: DA-backed GitHub storage (AEMXSC/XSCTeamSite)
- Auth: GitHub PAT for content operations, Claude API key for AI
compass/
├── index.html # Single-page app entry
├── app.js # Core application logic
├── app.css # Styles
├── ai.js # AI chat engine (streaming, tools, page context)
├── aem-knowledge.js # Distilled AEM skill intelligence
├── xsc-playbook.js # XSC pre-sales playbook
├── customer-profiles.js # Customer profile management
├── mcp-client.js # MCP connector framework
├── github-content.js # GitHub content API
├── da-client.js # Document Authoring client
├── site-detect.js # AEM site detection
├── known-sites.js # Known AEM site registry
├── governance.js # Brand governance engine
└── docs/ # Documentation
Adobe XSC Team — Experience Success Center
v0.1 — Project Compass