AgenticMail — collab record (Tam + Echo, 2026-10-19)

Both lanes agree — and Echo added the decisive evidence

Eve's own docs list first-party support for every channel AgenticMail offers — Twilio, Resend, Gmail, Novu, WhatsApp, Telegram, Slack, Discord, Teams, Messenger + six SMS providers. Verified live by me on eve.dev/integrations (channel keywords present in the rendered docs page tonight; 404s on eve.ch mirror path did not survive the check).

Adopting AgenticMail = a parallel implementation of channels EVE already carries (the exact "restated fact" defect class the V3 architecture exists to remove).

My earlier study summary (supplementary, all still standing)

• Repo shape verified: ★239, MIT, TypeScript, pre-1.0 churn (v0.9.x band), last substantive commit Sep-13; enterprise-grade not yet (Dependabot bumps dominate).
• Its runtime = TWO mail estates in one: Stalwart (containerized SMTP/IMAP/HTTP) + SQLite store + PM2 dispatcher + optional external Gmail relay — heavy for a node that already carries the tenant's O365 plan and EVE.
• Skim-scan vs Echo's sharper lens: Echo also caught what I sampled at and did NOT — the SECURITY.md still supports "0.2.x" while the product ships 0.9.54 = disclosure process stale → the security posture is unverified DESPITE the shape (outbound guard, spam filter, DKIM) measuring as soundly designed. Sharp lens and good catch.
• SMS/Voice glue is consumer-app-facing (Google Voice bridges etc.) = ToS-fragile; our business SMS lane stays Twilio-family when a customer demands it (Mira's cost rule: avoid SMS by default).

Where the value transfers (Echo's list, identical to my conclusions)

1. Outbound guard + human approval (Stripe-like) → EVE HIL pattern — no code, the design stance.
2. ask_operator (HIL on a live call, re-dial on disconnect) → good phone-UX for our later voice lane.
3. Wake/coalescing for agent-to-agent threads (multi-agent mail coordination) → relevant only if we ever adopt mail-as-fleet-transport, which we do NOT (A2A stays the fleet plane).

Register reconciliation (Mira + Min — clean)

• My agenticmail-eve-study.md (04c673c) + Echo's agenticmail-study-2026-10-19.md (b4306c7) both exist in Projects/vhost/research/ — two docs, same conclusion, complementary evidence. Recommendation for Min: a one-line cross-link note on each pointing to the other; merging coincidental bodies into one would throw away good footnotes.
• The package's 3-in-1 architecture spec needs NO change; the EVE HIL design takes the pattern as a single bullet tonight.

🎯 Tam's Read: two research lanes, one verdict — the EVE package loses nothing by declining AgenticMail because it already hosts every channel that packet offered; the fleet takes the approval-gate posture (already our discipline) and the name goes into the register as a future re-check ONLY if a tenant asks for inbound email/voice (Echo's named trigger, adopted verbatim).