Architecture: What notification approach for inter-agent communication?

Documentation for Sessions

🏗️ Architecture: What notification approach for inter-agent communication?

Date: 2026-02-06
Time: 23:58:33 UTC
Session: editor2-97.20
Agent: architecture
Confidence: 95%
Status: ✅ Documented


📋 Decision Context

Question:

What notification approach for inter-agent communication?

Task: Design inter-agent notification system


🎯 Decision Made

Chosen:

Layered approach: (1) Existing WebSocket infrastructure (backbone), (2) MCP notification injection (coding agent awareness), (3) Native macOS osascript notifications (human awareness), (4) VSCode extension notifications (IDE-level). Rejected Twilio (agents can't read SMS), Firebase FCM (overkill, wrong target), and webhooks (redundant with existing WS)

Options Considered:


💭 Reasoning

Twilio/FCM are designed for human-facing notifications on phones — AI agents can't read SMS. The existing WebSocket infrastructure is already the right transport layer. The real gap is the MCP protocol limitation (pull-only, no server-to-LLM push). Solved with notification injection on tool responses + native OS notifications for human awareness. No external service dependency needed.


📁 Files Involved


🔗 Related


Auto-generated by MCP Smart Consolidation