Universal Zadarma integration for Twenty CRM — product strategy & roadmap

[Nik1125]

Why

Twenty Apps graduated to production. Goal: turn this bespoke Algeness integration into the universal, official-quality Zadarma integration for Twenty CRM — and a credible partnership/acquisition target for Zadarma.

Headline insight: the app is already far more universal than it looks. Most Algeness-specific behaviour lives in n8n (naming, SMS classifier, templates), not in app code. The in-app coupling is minimal and mostly optional — so the cleanup is smaller than it first appears.

Acquisition vs partnership — realistic framing

Acquisition is a north star, not a direct target. A small integration alone does not trigger a buyout. Realistic near-term win: official partnership with Zadarma (recommended integration, co-marketing, revenue-share, roadmap access). Acquisition becomes plausible only after real traction + a premium feature Zadarma would rather buy than build (SMS messenger, click-to-call, generic AI).

Universal core (keep / strengthen)

Needed by every Zadarma + Twenty user:

• Call/SMS ingest (webhooks) → callLog / smsLog, auto-linked to Person
• Phone→Person matching (last-9-digit, 3-pass)
• Transcripts + recordings (link-only, GDPR)
• Send-SMS from Person panel
• Sync-calls (history pull), orphan-rescan (data-integrity repair)

In-app coupling to strip / generalize

• Retell hardcoded (AI_EXTENSIONS, callerType=AI, aiCost, correlationId) → optional AI layer, silent when AI_EXTENSIONS empty. /call-enrichment is already vendor-agnostic.
• TeamSale sync → keep optional (Zadarma's own CRM — a plus for acquisition, not a minus).
• Naming conventions / SMS-template enums → already n8n-side, not in app. OK.

Gaps to product-quality

Gap	Action
Reliability: uncaught throw → HTTP 500 (15 handlers)	try/catch all handlers → 200 {ok:false}
Test coverage	vitest on HMAC verify, phone-match 3-pass, webhook parse (all payload shapes)
Click-to-call (missing)	Zadarma /v1/request/callback/ — universal killer feature
Docs	setup-guide for generic user, webhook troubleshooting
Distribution	Twenty Marketplace + Zadarma integrations directory

Roadmap

Phase 1 — Universal core (clean public app)

• Fork to separate clean project; Algeness → private overlay
• Retell/Algeness specificity → optional layer (AI off by default)
• try/catch all handlers (500 → 200 ok:false)
• Tests on core (phone-match, HMAC, webhook parse)
• README + setup-guide for generic user
• Twenty Marketplace publish

Phase 2 — Distribution + partnership

• Submit to Zadarma integrations directory
• Submit to Twenty featured apps
• Collect installs / feedback
• Approach Zadarma with traction → official partnership talks

Phase 3 — Premium (path to acquisition)

• SMS messenger (threaded, per-Person + global inbox — after FC 2.20 regression)
• Click-to-call
• Generic AI analysis (any vendor — premium tier)
• With traction + premium → acquisition talks

Premium differentiators (what would actually attract Zadarma)

1. SMS messenger — threaded two-way SMS in a Talqui-style window. Fills a gap Twenty doesn't cover natively. Feasibility: Twenty FC is sandboxed iframe (#22588) — rebuild after Phase-2 FC regression; the model is maturing (runAgent, KV-store, i18n).
2. Click-to-call from CRM.
3. Generic AI analysis (not Retell-locked).

Full reference

docs/UNIVERSAL-PRODUCT-STRATEGY.md — companion doc with full depth (EN TL;DR + Russian working text + acquisition framing + conventions for the universal fork). Lives on the universal-core branch.

Discussion

This is the anchor issue for the universal-product direction. Separate branch + deep cleanup to follow. Input wanted on:

• Scope of the Phase 1 cut (what's truly core vs optional)
• Messenger feasibility priorities (which FC-2.20 blockers first)
• Partnership outreach strategy (who at Zadarma, what to lead with)

Related: #71 (Twenty 2.6→2.20 upgrade — the platform foundation this builds on).