Kasha 2026 Q1 (&Q2) Project Plan

Q1-Q2 2026 · SMB · Enterprise · CSP · Private Banking · 6 sprints, Feb — Jun

Kasha Repos

Heidi Monorepo

Microservices

5,650

Heidi Tests

2,372

Kasha Test Files

Sprints

Executive Summary

Four product lines, one platform. Kasha SMB — all-in-one SaaS for service businesses (POS, booking, CRM, invoicing). Kasha Enterprise — Voucherised Payments for one-to-many disbursements with near-zero barriers (no sender KYC). Kasha CSP — CSP platform (Heidi) that lets Corporate Service Providers offer co-branded SMB + Enterprise to their formed entities. Kasha Private Banking — bank product reselling as licensed agents.

Enterprise and CSP are B2B2B feeders for SMB. Every Enterprise payee is a potential SMB merchant. Every CSP-formed entity gets co-branded access to the entire ecosystem. Private Banking deepens lock-in across all product lines.

The strategic move: Migrate UK CSP operations (Companies House, HMRC) from Kasha into Heidi. Heidi becomes the multi-jurisdiction CSP layer (CH + UK + future markets). Kasha Platform provides SMB + Enterprise + Private Banking to all customers — direct, CSP-provisioned, and Enterprise-fed.

Kasha Platform — Current State

33 repos: 18 NestJS services, 10 Next.js apps, 4 shared libs
v2.0 Phases 1-3 done (NestJS 11, Next.js 15, React 19, TS 5.8)
Phase 4+ in progress: code quality debt, globalization (3/10), security hardening

Heidi (KashaCH) — Current State

Full-stack monorepo: React 19 + Vite 7 + Express + Drizzle ORM + PostgreSQL
224 test files, 5,650 passing tests
Swiss company formation: AG, GmbH, Foundation entity types
Multi-tenant CSP model (platform mode + CSP mode)
Modules: companies, beneficial owners, obligations, tasks, billing, KYC (SumSub), AI agent
Live at company-formation.letscodenow.app

Two-System Architecture

Heidi owns the client lifecycle. Kasha provides the tools. The integration point is tenant provisioning.

CSP Layer — Heidi

Company Formation Compliance Obligations KYC / AML Billing AI Agent

Jurisdictions

Switzerland (AG, GmbH, Foundation) UK (Ltd, Plc) ← migrating in Future markets

↓ Auto-provision tenant via API ↓

SaaS Layer — Kasha Platform

API Gateway

Auth CRM Invoicing Banking Booking E-com Search Notifications

PostgreSQL MongoDB Redis Elasticsearch

Integration Flow

CSP forms company in Heidi (any jurisdiction)
Heidi calls Kasha API to provision a new tenant (company, owner, auth)
Kasha returns credentials — client gets SaaS access (booking, invoicing, CRM, etc.)
Ongoing sync: company status, billing, compliance flags flow between systems
SSO linkage: CSP staff can access client's Kasha dashboard from Heidi

Heidi CSP Platform

Multi-tenant CSP platform. Currently Swiss-only. Expanding to UK jurisdiction.

Component	Stack
Frontend	React 19, Vite 7, TailwindCSS 4, Radix UI, React Query
Backend	Express.js, TypeScript, Drizzle ORM
Database	PostgreSQL
Testing	Vitest, Testing Library (224 files, 5,650 tests)
AI	Anthropic SDK for AI agent capabilities
External	SendGrid, AWS S3/GCS, SumSub (KYC)

UK CSP Migration Scope

Add Ltd and Plc entity types alongside AG/GmbH/Foundation
Companies House API integration (formation, filings, director updates)
HMRC registration integration (VAT, PAYE, Corporation Tax)
UK-specific compliance obligations (confirmation statements, annual accounts)
UK beneficial owner reporting (PSC register)
Multi-currency billing (CHF + GBP)
Move UK CSP concerns OUT of Kasha Platform into Heidi

AI Development Cycle

Three AI agents across both platforms. TLC methodology on both Kasha (Mocha/Vitest) and Heidi (Vitest).

Agent	Role	Capabilities
Claude	Primary (70%)	Architecture, code-gen, review, refactor, debugging
Codex	Secondary (15%)	Code-gen, consensus review with Claude
Gemini	Creative (15%)	Design, vision, prototyping, UI mockups, brainstorming

Architectural Decisions

Key technical decisions recorded for the project. These inform sprint planning and constrain implementation choices.

ADR-001: REST stays, GraphQL BFF later

Status: Accepted When: Q2-Q3 (with mobile)

Context: 18 NestJS microservices with REST APIs, 10 Next.js frontends, API Gateway on NestJS 10. React Native mobile apps planned for Brazil & Switzerland. Question raised: should we migrate to GraphQL?

Decision: Do NOT migrate existing REST APIs. Add a thin GraphQL BFF (Backend-for-Frontend) layer when building the React Native mobile apps (Brazil GTM Phase 2, Q2-Q3 2026).

React Native App

↓

GraphQL BFF (new, thin, @nestjs/graphql + Apollo)

↓

REST API Gateway (existing, unchanged)

↓

18 NestJS Microservices (existing REST, unchanged)

Rationale:

No rewrite risk: 18 services, every controller, every route, every test would need rewriting. That's a quarter of work, not a sprint.
Gateway not ready: lr-api-gateway is still on NestJS 10 — upgrade it first before adding complexity.
External APIs are REST: Xero, QuickBooks, HMRC, Companies House, Transfero, Bexio, Amazon, Google Shopping — all REST/OAuth. Backend stays REST regardless.
Two platforms: Heidi is Express + REST. Diverging API paradigms makes the Heidi ↔ Kasha integration harder.
Real bottlenecks are elsewhere: 158 hardcoded GBP refs, 3/10 globalization, zero accounting integrations, 18 empty catch blocks.
BFF gives the upside without the risk: Mobile gets flexible queries, existing APIs stay untouched, NestJS supports this pattern natively.

Consequences:

Existing Next.js frontends continue using REST (no change)
React Native apps use GraphQL BFF for optimized mobile queries
BFF aggregates across multiple REST services per screen
Heidi ↔ Kasha integration remains REST-based (tenant provisioning API, sync)
Future option: if GraphQL BFF proves valuable, web frontends can optionally migrate to it incrementally

ADR-005: Middleware Connector Pattern (Platform Unity)

Status: Accepted & Proven KYC: Built

Context: Two codebases — KashaCH (Heidi, Vite+Express monolith) and Kasha Platform (43-repo NestJS microservices). Both need shared services (KYC, payments, CRM, compliance). Building each twice creates duplication and divergence.

Decision: Bridge platforms via Middleware Connectors — thin, standalone NestJS REST APIs that own a single business domain. Both platforms consume the same REST API. Zero duplication.

KashaCH (Heidi) Kasha Platform

↓ HTTP REST ↓

Middleware Connector (NestJS Gateway)

↓ uses ↓

Shared Library (@kasha/{domain}-service)

↓ adapters ↓

Shared PostgreSQL + External Providers

Three-repo structure per domain:

Shared Library (kasha-{domain}-service) — provider-agnostic business logic, store interfaces, DTOs, events
Middleware Connector (kasha-middleware-{domain}-connector) — NestJS REST gateway, TypeORM adapters, Docker + CI/CD
Platform Consumer (lr-{domain}-service) — calls connector API or imports library directly

Proof of concept (KYC): kasha-kyc-service + kasha-middleware-kyc-connector — 11 endpoints, 10 DB entities, SumSub + Mock providers, deployed via Docker.

Next connectors: Payments (next up), CRM (planned), Compliance (planned).

Full case study: Platform Unity — KYC Middleware Connector

ADR-006: Domain-Specific Shared Libraries

Status: Accepted & Proven @kasha/kyc-service: Built

Context: With two platforms consuming the same business domains, logic must be centralised to prevent divergence.

Decision: Extract all shared business logic into provider-agnostic libraries published on GitHub (@kasha/{domain}-service). Libraries define orchestrator, store interfaces, DTOs, providers (real + mock), events, and config. No database opinions — store interfaces are abstract. The middleware connector implements them with TypeORM; other consumers can use any ORM.

Key pattern: Mock providers auto-detect from environment variables. If SumSub token is missing or "dev-placeholder", library automatically falls back to MockKycProvider — zero config, zero feature flags.

Sprint 1: Code Quality & Security

Feb 10 — Mar 2 AI Lead: Claude + Codex

Kasha Platform — Code Quality

Fix 18+ empty catch blocks across lr-kasha-builder-library
Reduce @ts-ignore / any by 50% (200+ → 100)
Refactor 6 files exceeding 2,000 lines
Resolve 35+ TODO/FIXME items (triage: fix or remove)
Standardize error handling pattern across backend services

Kasha Platform — Security

Fix cookie security (secure: false hardcoded)
Add environment variable validation to all services
Review 13 query construction patterns for SQL injection risks
Remove hardcoded configuration values (ports, URLs, secrets)

Kasha Platform — Testing Foundation

Enable 12+ skipped tests in lr-chat-widget-ui
Add test framework to lr-fe-components (currently 0 tests)
Expand lr-audit-service (3 → 30+ tests — compliance logging is critical)
Expand lr-search-service (2 → 20+ tests)

Heidi Foundation

Design tenant provisioning API contract (Heidi → Kasha) — see Integration Spec
Audit Heidi codebase for UK expansion readiness
Design multi-jurisdiction schema (entity types, obligations per jurisdiction)

Sprint 2: Globalization & Heidi UK

Mar 3 — Mar 23 AI Lead: Claude (architecture) + Codex (bulk ops)

Kasha Platform — Globalization

Upgrade lr-api-gateway from NestJS 10 → 11
Extract hardcoded GBP/£ to configuration (158+ locations)
Replace UK postcode regex with international address validation
Abstract Companies House → pluggable company registry (move CSP logic to Heidi)
Abstract banking integration (Railsr → pluggable PSP layer)
Multi-timezone support (remove Europe/London hardcoding)
TypeORM standardization to 0.3.28 across all services
Tailwind CSS 3 → 4 migration across 10 frontend apps

Kasha Platform — Integration

Build tenant provisioning API endpoint (receives calls from Heidi)
Implement shared API key auth for service-to-service calls
Set up integration test environment (Heidi staging ↔ Kasha staging)

Heidi UK Jurisdiction

Add Ltd and Plc entity types to company schema
Companies House API integration (formation, filings, directors)
UK compliance obligations engine (confirmation statements, annual accounts)
UK PSC (Persons of Significant Control) register integration
Multi-currency billing (CHF + GBP)
Tenant provisioning: call Kasha API on company formation

Sprint 3: Accounting Integrations

Mar 24 — Apr 13 AI Lead: Claude (API) + Gemini (UI)

Kasha Platform

Xero integration: OAuth 2.0, invoices, contacts, payments, bank txns
QuickBooks Online: invoices, customers, payments, products, tax codes
Sync engine with conflict resolution and retry logic
Admin UI for connection management and monitoring

Heidi Accounting Link

Auto-connect Xero/QB for newly provisioned Kasha tenants
Billing data sync: Heidi CSP fees → Kasha invoicing
Bexio integration planning (Swiss accounting, P3)

Sprint 4: MTD + Mass Payout v2

Apr 14 — May 4 AI Lead: Claude (compliance) + Codex (tests)

Kasha Platform — Making Tax Digital

New lr-tax-service: HMRC OAuth 2.0, VAT obligations, submission, fraud prevention headers
New lr-tax-portal-ui: VAT return review & submission UI
HMRC sandbox testing & production credentials
Digital record keeping compliance

Kasha Platform — Mass Payout v2

Builds on existing production system (CSV upload, validation, multi-currency, approval workflow, BullMQ processors).

Phase 1 — Contractor Profiles: OrganizationPayoutProfile in lr-company-service (tax residency, VAT status, self-billing consent), W-9/W-8BEN upload + validation
Phase 2 — Tax Rules Engine: New lr-tax-rules-service (stateless, 11 initial rules: UK, EU, US, UAE), VAT/treaty rate tables, withholding calculation in payout pipeline
Phase 3 — Auto-Invoice: Generate invoice from payout batch item, tax treatment from rules engine, link to MassPayout entities, webhook delivery

Open questions: Invoice ownership (payer vs payee?), self-billing jurisdiction rules, credit-note flow for failed payouts — see spec §7, §14

Heidi Tax Integration

HMRC registration flow for UK companies formed via Heidi
VAT registration status sync (Heidi ↔ Kasha tax service)
Swiss tax filing obligation tracking (cantonal differences)

Sprint 5: Marketplace & Channels

May 5 — May 25 AI Lead: Gemini (UI) + Claude (APIs)

Kasha Platform

New lr-marketplace-ui: public catalog, merchant storefronts, checkout
New lr-channel-manager-service: product syndication
Amazon SP-API + Google Shopping initial integration

Heidi CSP Marketplace

CSP can offer marketplace access as a service package add-on
Auto-enable marketplace features on tenant provisioning based on package tier

Sprint 6: Advanced Features

May 26 — Jun 15 AI Lead: Claude (finance) + Gemini (dashboards)

Kasha Platform — Receivables Financing (Foundation Only)

Sprint 6 is the kickoff of a 16-week effort. Full spec has 5 phases — we target Phase 1 here.

Phase 1 only: New lr-financing-service — data models for financeable_receivable, financing_advance, risk_profile, pricing_tier
Eligibility engine: Basic rules, daily rate pricing, manual approval only
Switzerland model: T+2/T+1 settlement — anticipation based on minimum 3-month relationship; capital availability tied to payment history and volume consistency
Brazil model: Anticipation on known payout schedules (e.g., Carrefour 180-day terms); confirmed receivable from anchor buyer
Key principle: Receivables purchase (not lending) — regulatory distinction matters

Phases 2-5 continue into H2: Mass payout integration, card processing integration, risk scoring + auto-approval, auto-advance + enterprise API

Kasha Platform — Mass Payout v2 (cont.)

Phase 4 — Voucher & Breakage: Extend claim flow with 1-year voucher expiration, reminder scheduler, breakage revenue recognition
Phase 5 — Reporting: 1099/EC Sales List generation, enterprise payout analytics dashboard
Sage Business Cloud integration (UK/EU)

Heidi Future Jurisdictions

Architecture review: what's needed for jurisdiction #3 (e.g., EU/Dubai)
CSP analytics dashboard: cross-jurisdiction overview
AI agent enhancements: jurisdiction-aware compliance recommendations

Integration Spec: Heidi ↔ Kasha

The integration is REST-based (see ADR-001). Three integration points:

1. Tenant Provisioning API

Heidi calls Kasha when a CSP forms a company. Designed in S1, built in S2.

Aspect	Detail
Endpoint	`POST /api/v1/tenants/provision` on Kasha API Gateway
Auth	Service-to-service API key (rotatable, per-environment)
Payload	Company name, entity type, jurisdiction, owner (name, email), package tier
Response	Tenant ID, admin credentials, dashboard URL, enabled modules
Idempotency	Idempotency key in header — retry-safe, no duplicate tenants
Failure	Heidi retries with exponential backoff (3 attempts). Dead letter queue for manual review.

2. Ongoing Data Sync

Data	Direction	Frequency	Mechanism
Company status (active/dormant/dissolved)	Heidi → Kasha	On change	Webhook
Billing events (invoice paid/overdue)	Heidi → Kasha	On change	Webhook
Compliance flags (KYC expired, obligation due)	Heidi → Kasha	On change	Webhook
Usage metrics (bookings, invoices, transactions)	Kasha → Heidi	Daily batch	Scheduled API pull

3. SSO / Cross-Platform Access

Mechanism: Signed JWT with short expiry (5 min), generated by Heidi, validated by Kasha
Flow: CSP staff clicks "Open Kasha Dashboard" in Heidi → Heidi generates signed token → redirect to Kasha with token → Kasha validates & creates session
Scope: Read-only initially. CSP sees client's dashboard but cannot modify billing or auth settings.

Test Plan

Testing is not a separate sprint — it's woven into every sprint. Each sprint has specific test targets.

Sprint	Kasha Test Targets	Heidi Test Targets
S1	lr-audit-service (3 → 30), lr-search-service (2 → 20), lr-fe-components (0 → framework), enable 12 skipped tests	Multi-jurisdiction schema tests
S2	lr-auth-service MFA/password reset paths, lr-banking-service payment/refund paths, tenant provisioning endpoint tests	UK entity type tests, Companies House integration tests, provisioning call tests
S3	Xero/QB sync engine tests (mock responses), conflict resolution tests	Auto-connect accounting tests, billing sync tests
S4	lr-tax-service HMRC submission tests (sandbox), lr-tax-rules-service rule evaluation (11 rules), contractor profile validation	HMRC registration flow, VAT sync, cantonal obligation tests
S5	Marketplace listing/search/checkout tests, channel manager syndication tests	Package tier provisioning tests
S6	lr-financing-service eligibility/pricing tests, voucher/breakage lifecycle tests	CSP analytics query tests, jurisdiction config tests

Critical Coverage Gaps (from Backlog)

Service	Current	Risk	Sprint
lr-audit-service	3 tests	Critical — compliance logging	S1
lr-search-service	2 tests	Critical — search indexing	S1
lr-banking-service	Partial	Critical — payment/refund paths	S2
lr-auth-service	Partial	Critical — MFA, password reset	S2
lr-kyc-service	Partial	Critical — approval/rejection	S2
lr-fe-components	0 tests	Med — shared UI lib	S1
lr-notifications-service	9 tests	Med — email/SMS delivery	S3
lr-crm-service	11 tests	Med — contracts, clients	S3

UK CSP Data Migration

Moving UK CSP concerns from Kasha to Heidi. This is code + data + integration logic.

What Moves

Component	From (Kasha)	To (Heidi)	Sprint
Companies House API client	lr-company-service	Heidi companies module	S2
UK entity type logic	lr-company-service	Heidi formations module	S2
HMRC registration logic	lr-company-service	Heidi compliance module	S4
UK compliance obligations	Scattered / manual	Heidi obligations engine	S2
PSC register integration	None (new)	Heidi beneficial owners	S2

Migration Strategy

Parallel run: Both systems active for 1 sprint (S2). Heidi builds UK features while Kasha still serves existing UK clients.
Cutover: End of S2 — Heidi takes over UK CSP operations. Kasha's Companies House client deprecated (kept as fallback for 1 sprint).
Rollback: Feature flag per jurisdiction in Heidi. If UK breaks, revert to Kasha's Companies House client.
Data: Existing UK company data (if any) exported from Kasha PostgreSQL, imported to Heidi via migration script. Tested in staging first.

Brazil GTM Track

Brazil is a H2 2026 launch. The Feb-Jun sprints lay the foundation (globalization, pluggable architecture). Dedicated Brazil work starts Q3.

Item	Dependency	When
Globalization (multi-currency, i18n framework)	None	S2 (in plan)
Pluggable banking layer (Railsr abstraction)	S2 globalization	S2 (in plan)
Transfero PSP integration (Pix, BRZ stablecoin)	Pluggable banking layer	Q3 — post-sprint-6
i18n content (pt-BR translations)	i18n framework from S2	Q3
React Native mobile apps	GraphQL BFF (ADR-001)	Q3
Luma AI (pt-BR)	i18n, Transfero live	Q3-Q4
WhatsApp bot integration	Luma AI pt-BR	Q4
Vertical landing pages	Marketplace (S5)	Q3
Receivables financing (Brazil model)	lr-financing-service (S6)	Q4

Brazil critical path: S2 globalization → S2 pluggable banking → Q3 Transfero → Q3 mobile apps → Q3 MVP launch (São Paulo)

Switzerland GTM Track

Two sub-tracks: CSP (Heidi) is in-plan. Domestic (Kasha for Swiss SMEs) overlaps with globalization and extends into H2.

CSP Track (in-plan)

Already covered in sprints: UK jurisdiction (S2), tenant provisioning (S1-S2), compliance automation (S2-S4), CSP analytics (S6).

Domestic Track (partial overlap + H2)

Item	Dependency	When
TWINT payment optimization	Adyen integration (exists)	S2 — part of globalization
Multi-currency engine (CHF)	S2 GBP extraction	S2 (in plan)
Bexio accounting integration	Sync engine from S3	Q3 — uses same sync engine as Xero/QB
Swiss QR-bill (QR-Rechnung)	lr-invoicing-service	Q3
Luma AI in DE/FR/IT	i18n framework from S2	Q3
CSP onboarding pipeline	Heidi tenant provisioning (S2)	Q3 — white-glove, entity migration tools
Local marketplace (Zürich)	Marketplace from S5	Q4
CH anticipation engine	lr-financing-service (S6)	Q4

Delivery Timeline

Feb

Mar

Apr

May

Jun

Kasha Platform

Code Quality

Globalization

Xero + QuickBooks

MTD / Tax

Mass Payout v2

Marketplace

Financing

Heidi CSP

UK Jurisdiction

Tenant Provision

Tax Integration

Future Jurisdictions

GTM Tracks

Brazil Foundation

Brazil Launch (Q3)

CH CSP (in-plan)

CH Domestic (Q3)

Ongoing

Testing + Quality

Operations

Risk Matrix

Risk	Level	Impact	Mitigation
Heidi-Kasha integration complexity	High	Sprint 2-3	Design API contract early (S1); versioned API; integration tests
UK CSP migration scope	High	Sprint 2	Phase UK features; CH stays stable; feature flags per jurisdiction
HMRC API changes	High	Sprint 4	Sandbox-first; abstract HMRC client; monitor changelog
Globalization scope creep	High	Sprint 2	Automated replace for obvious patterns; phase into tiers
Data model divergence	Med	Ongoing	Shared schema definitions; sync validation; contract tests between systems
Third-party rate limits	Med	Sprint 3	Mock responses for tests; dedicated sandbox accounts
Financing licensing (BR/CH)	Med	Sprint 6+	Receivables purchase (not lending) model; legal review per jurisdiction; start with manual approval only
Multi-tenancy readiness	Med	Sprint 2	Audit current Kasha tenant isolation; verify data partitioning before provisioning API goes live
Brazil PSP dependency	Med	Q3	Transfero partnership must be signed before Q3 dev starts; fallback: Adyen Brazil
AI team capacity	Low	Velocity	TLC guardrails; consensus review; reusable prompts

KPIs & Success Metrics

Metric	Current	Target (Jun)
Kasha test coverage	Varies	80%+ critical services (auth, banking, invoicing, audit, kyc)
Heidi test count	5,650	7,000+ (UK jurisdiction tests)
Empty catch blocks	18+	0
TODO/FIXME items	35+	<10 (resolved or tracked as issues)
SQL injection risk patterns	13 unreviewed	0 (all reviewed + fixed)
Globalization score	3/10	8/10
Heidi jurisdictions	1 (CH)	2 (CH + UK)
Tenant auto-provisioning	None	Live (Heidi → Kasha, with SSO)
Accounting integrations	0	3 (Xero, QuickBooks, Sage)
HMRC MTD compliance	None	100% (VAT submission live)
Marketplace channels	0	3 (internal + Amazon + Google)
Integration test env	None	Live (Heidi staging ↔ Kasha staging)

Operations & Infrastructure

33 Kasha repos + 1 Heidi monorepo. Operational maturity must match feature velocity.

CI/CD Pipeline

System	Current	Target
Kasha	GitHub Actions per-repo (lint + test + build)	Add integration test stage, coverage gates (80%+ critical), deploy-on-merge to staging
Heidi	Monorepo CI (lint + 5,650 tests + build)	Add UK jurisdiction test suite, E2E provisioning test, coverage gates
Cross-system	None	Integration pipeline: Heidi staging → Kasha staging provisioning test (nightly)

Environments

Environment	Purpose	Status
Local	Developer workstation (Docker Compose)	Exists — needs unified compose for Heidi + Kasha
Dev	Feature branch deployments, manual QA	Kasha: exists. Heidi: exists. Cross-system: Missing
Staging	Pre-production, integration testing, UAT	Build in S2 — both systems on staging with provisioning live
Production	Live traffic	Heidi: company-formation.letscodenow.app. Kasha: production cluster.

Monitoring & Observability

Logging: Structured JSON logs across all NestJS services. Heidi: Winston. Kasha: mixed (standardize in S1).
Error tracking: Sentry integration for both platforms (existing in Heidi, add to Kasha critical services in S1).
Health checks: /health endpoint on every NestJS service. Uptime monitoring via external ping.
Alerting: Sentry alerts for error spikes. Webhook to Slack for deployment failures and provisioning errors.
Metrics (H2): Prometheus + Grafana for request latency, queue depth, provisioning success rate, financing pipeline metrics.

Deployment Strategy

Kasha: Containerized (Docker). Individual service deploys. Blue-green for critical services (auth, banking, invoicing).
Heidi: Single container deployment (monorepo). Zero-downtime rolling updates.
Database: Migrations run as part of deploy pipeline (TypeORM for Kasha, Drizzle for Heidi). Rollback scripts required for every migration.
Feature flags: Jurisdiction-level flags in Heidi (per S2). Service-level flags in Kasha for new features (financing, marketplace).
Secrets: Environment variables via CI secrets. No hardcoded secrets (audit in S1). Rotate API keys quarterly.

Open Decisions

Items that need a decision before or during the sprints. Each has a deadline and a decision owner.

Decision	Options	Deadline	Impact
Brazil BACEN licensing Decided	Operate under Transfero's license. No own payment institution application. Pass-through model — Transfero holds the BACEN license, we build the SaaS layer on top.	—	High — architecture: pass-through via Transfero, no own ledger
Swiss FINMA considerations Decided	Receivables purchase (factoring), not lending. No FINMA banking license required. Same model as Brazil — buy the receivable, don't extend credit.	—	High — no license needed; product is invoice factoring, not credit
Transfero PSP contract Decided	Dual PSP: Transfero + Adyen Brazil. Transfero for Pix, BRZ stablecoin, BACEN license. Adyen for international cards, fallback. Pluggable banking layer (S2) supports both.	—	High — dual-PSP de-risks Brazil GTM; pluggable layer in S2
Mobile strategy Decided	In-house native. React Native (with GraphQL BFF per ADR-001) or alternative native strategy. No agency outsourcing. AI-assisted dev (Claude + Codex).	—	Med — React Native vs. alternative native TBD by end of S4
Financing capital source	Own balance sheet vs. bank partner vs. institutional fund vs. tokenized (BRZ)	Before S6	Med — affects risk model, capital requirements, and regulatory path
Kasha multi-tenancy audit	Full audit + fix vs. audit + document gaps vs. defer	S1 (blocking for provisioning API)	High — provisioning API cannot go live with weak tenant isolation
Pricing model: CSP packages Decided	Platform Fee + Formation Fees + Merchant-Pays + CSP Revenue Share. No per-entity monthly fees. See CSP Pricing Model below.	End of S2	Med — affects package tier field in provisioning API and Heidi billing
Marketplace commission model Proposed	Subscription-inclusive + tiered commission + channel syndication fee. See Marketplace Pricing below.	Before S5	Med — affects lr-marketplace-ui checkout and revenue recognition

Phase: Licensing Framework, Routing & MSA

Discussion Phase Status: Scoping

Multi-license, multi-partner architecture determining how every payment flow, payout, IBAN issuance, and cross-border transaction is routed. Two own licenses (Swiss SRO, Canadian MSB+RPAA), three partner licenses (Malta PI with Latvian IBAN/Target2, Transfero Brazil, Adyen global), five markets, five product lines.

Full discussion document: License Framework, Routing & MSA →

Covers: own & partner license details, routing tables per market (CH, BR, EU, CA, UK), fund flow diagrams, safeguarding, settlement rails, MSA structure with product-specific schedules, data processing per jurisdiction, and 9 open questions requiring resolution.

Customer Acquisition: Team Reality

No sales team. No BDRs. No field reps. A nimble marketing team, a few devs, multi-tasking team members, and AI that produces content at 100x human speed. This is how we actually acquire customers across both markets.

Role	People	Acquisition Contribution
Product / Engineering	3-4 devs	Build hooks that sell themselves: free TWINT/Pix, booking pages, Luma AI. Product IS the acquisition strategy.
Marketing	1-2 people	Campaign strategy, brand voice, creative direction. AI handles volume — humans handle taste.
AI Agents	Claude, Codex, Gemini	Content factory: 50-100 pieces/week per market — SEO, social, email, landing pages, ad copy, all languages.
Founder / Leadership	Multi-tasking	CSP relationships (3-5 total), strategic partnerships (Bexio, trade associations), 1-2 events/year.
Everyone else	Multi-tasking	Customer success, onboarding support, feedback loops. No dedicated sales — every touch point is product-led.

What we don't do: Cold call SMEs. Attend 10+ events/year. Hire a sales team in 2026. Run brand campaigns. Build a partner/reseller channel. Sales team consideration starts at 1,000+ merchants, not 300. Every CHF/BRL spent has a trackable conversion.

AI Content Factory

Our unfair advantage. While competitors hire agencies and wait 4 weeks for a campaign, we generate, test, and iterate daily.

Channel	Weekly Output	AI Does	Human Does
Instagram / TikTok	15-40 posts per market	Captions in DE/FR/IT/EN/pt-BR, hashtags, A/B variations, Reels scripts, carousel copy	Visual direction, brand approval, scheduling
Google Local SEO	10-15 landing pages	Vertical × city pages: "Coiffeur Zürich buchen", "Salão de beleza agendar São Paulo"	Keyword strategy, conversion optimization
Email / WhatsApp sequences	Automated flows	Onboarding drip, activation nudges, upgrade prompts, win-back — personalized per vertical and language	Trigger logic, segmentation rules
Blog / Content	3-5 articles per market	"Wie akzeptiere ich TWINT?", "Como funciona antecipação de recebíveis" — SEO-optimized, local voice	Topic selection, fact-checking
Ad copy	20-30 variations per campaign	Google Ads, Meta Ads, A/B variants — Swiss German colloquial, Brazilian Portuguese	Budget allocation, performance review
Luma AI (in-product)	Continuous	Luma IS the demo AND the onboarding. Prospects talk to the product before signing up. Luma IS the sales rep on WhatsApp in Brazil.	Conversation design, dialect tuning

Volume beats perfection. A 50-person marketing team produces 10 polished pieces a week. Our AI factory produces 50-100+ pieces per market per week, tests them, kills the losers, doubles down on winners. We out-iterate, not out-spend.

Product-Led Growth

Every feature is an acquisition channel. No human needed at any stage.

Zero-Touch Acquisition Loops

Loop	How It Works	Market
Free TWINT / Pix	Merchant Googles "TWINT akzeptieren" / "receber Pix grátis". Finds SEO page. Downloads app. Accepts first payment in 5-10 minutes.	CH / BR
Booking Virality	Merchant shares booking link on Instagram / WhatsApp. Client books, sees "powered by Kasha". Their hairdresser / trainer asks "what's this?" — signs up.	Both
Luma Onboarding	New signup lands in app. Luma greets in local language, walks through setup, creates first invoice, configures payments. No human onboarding call.	Both
Referral	Built into app: "Empfiehl Kasha — CHF 20" / "Indique e ganhe R$30." WhatsApp/share link, automatic tracking, instant credit.	Both
Marketplace Pull	Consumer finds merchants on Kasha marketplace. Merchant A gets bookings. Merchant B sees it — signs up to be listed.	Both
Antecipação Lock-in	Merchant processes R$5K. App offers: "Antecipe seus recebíveis agora." Once they use working capital, switching cost rises dramatically.	BR
CSP Auto-Provisioning	CSP forms a GmbH through Heidi. Entity gets Kasha Corporate account instantly. Zero sales call. Zero onboarding.	CH
Channel Syndication	Merchant lists on Kasha → auto-syndicated to Mercado Livre / Amazon / Google Shopping. Kasha becomes the single pane.	BR

What Still Needs Humans

Activity	Who	Frequency
Approve AI content batches	Marketing	Daily (30 min)
Review ad performance, reallocate budget	Marketing	Weekly
Inbound support (Luma handles 80%+)	Team	As needed
CSP relationship management (3-5)	Founder	Monthly
Trade association partnerships	Founder	Quarterly
Finance events (STEP, SECA)	Founder	2x / year
Tune Luma AI per vertical	Product	Monthly

Funnel & Budget

Switzerland Funnel (month 6+, steady state)

Stage	Monthly Volume	Conversion
Impressions (AI content + ads)	15,000-25,000	—
Landing page visits	1,500-3,000	~10% CTR
Signups	60-120	~4% of visits
Activated (first transaction)	30-60	~50% of signups
Retained at month 3	20-40	~65% retention
Upgraded to paid	5-10	~20% of retained

Brazil Funnel (month 6+, steady state)

Stage	Monthly Volume	Conversion
Impressions (AI content + ads + WhatsApp)	50,000-100,000	—
Landing page / WhatsApp visits	5,000-10,000	~10% CTR
Signups	200-400	~4% of visits
Activated (first Pix payment)	100-200	~50% of signups
Retained at month 3	65-130	~65% retention
Upgraded to paid	15-30	~20% of retained

Acquisition Budget

Market	Monthly Ad Spend	AI Content Cost	Sales Salaries	Year 1 Total	Target Merchants	CAC
Switzerland	CHF 3-5K	CHF 0	CHF 0	~CHF 40-60K	300-500	~CHF 80-150
Brazil	BRL 10-15K	BRL 0	BRL 0	~BRL 120-180K	1,000	~BRL 50-80

CAC payback: Switzerland — CHF 80-150 CAC, CHF 49/mo Pro tier, payback in 2-3 months. Brazil — BRL 50-80 CAC, BRL 79/mo Pro tier, payback in 1-2 months. Traditional competitors with field sales teams spend 3-5x more per merchant. AI content + product-led growth is the moat.

Kasha CSP — Pricing Model (Decided)

CSPs pay for Heidi (their tool). CSP-formed entities get co-branded Kasha SMB + Enterprise access (min CHF 99/mo, no free tier). CSPs earn revenue share on all entity activity across all product lines. Enterprise and CSP are B2B2B feeders for SMB.

1. What the CSP Pays (Heidi Platform)

Fee Type	Amount	Notes
Platform Fee — Starter	CHF 299/mo	Solo fiduciaries, up to 25 entities, 3 users
Platform Fee — Professional	CHF 799/mo	Mid-size trust companies, up to 100 entities, 10 users, custom branding
Platform Fee — Enterprise	From CHF 2,500/mo	Large CSPs, 100+ entities, unlimited users, SLA, dedicated AM
Formation Fee — GmbH	CHF 390 (one-time)	Per entity formed through Heidi
Formation Fee — AG	CHF 590 (one-time)	Per entity formed through Heidi
Formation Fee — Foundation	CHF 790 (one-time)	Per entity formed through Heidi
Formation Fee — UK Ltd	CHF 190 (one-time)	Per entity formed through Heidi

No per-entity monthly fees. Platform fee covers Heidi access, entity management, compliance monitoring, obligation tracking, KYC/AML, document storage, and CSP dashboard.

2. What CSP Entities Pay (Co-Branded SMB + Enterprise)

CSP-formed entities get co-branded access to both Kasha SMB and Kasha Enterprise. They choose what they need. No free tier. Minimum CHF 99/mo — businesses paying CHF 25K+ to banks won't blink at this.

Product	Tier	Monthly	Includes
Kasha SMB (co-branded)	SMB Pro	CHF 99/mo (min)	Banking, invoicing (QR-bill), CRM, Bexio sync, Luma AI, compliance visibility
	SMB Business	CHF 149/mo	+ e-commerce, marketplace, multi-user, analytics, channel syndication
	SMB Premium	CHF 249/mo	+ API access, priority Luma AI, working capital, advanced reporting
Kasha Enterprise (add-on)	Enterprise	From CHF 499/mo	Voucherised Payments, mass payouts, no sender KYC, tax rules engine

Plus transaction fees (~1.5-2.5%), Enterprise payout fees (0.5-1.5%), marketplace commission (tiered), and working capital fees (3-5%).

Domestic Merchants (Kasha SMB — direct)

Tier	Monthly	Includes
Kasha SMB Free	CHF 0	TWINT, basic POS, booking page, Luma AI (acquisition hook)
Kasha SMB Pro	CHF 49/mo	+ invoicing, multi-staff, analytics, Bexio sync, QR-bill
Kasha SMB Business	CHF 99/mo	+ marketplace, channel syndication, anticipation, multi-location, API access

Same platform, different product surface. Direct SMB merchants get POS, booking, TWINT. CSP entities get co-branded SMB + Enterprise. Private Banking available to all.

3. What the CSP Earns (Revenue Share)

Revenue Stream	CSP Share	Example (40 entities)
Entity SMB subscriptions	15%	20 at CHF 99 + 15 at CHF 149 + 5 at CHF 249 = CHF 5,470/mo → CSP earns CHF 820/mo
Entity Enterprise subscriptions	15%	5 at CHF 499 = CHF 2,495/mo → CSP earns CHF 374/mo
Transaction & payout fees	10%	CHF 9,000/mo in fees → CSP earns CHF 900/mo
Working capital fees	5%	CHF 2,000/mo in fees → CSP earns CHF 100/mo
Private Banking commissions	10%	Referral commissions on bank products

Example: Mid-size CSP with 40 entities. CSP pays CHF 799/mo platform fee. All entities on SMB Pro min (CHF 99/mo), 15 upgraded, 5 on Enterprise. Revenue share back to CSP: ~CHF 2,194/mo — nearly 3x the platform fee. Heidi is a profit centre. Every entity can also access Private Banking products — additional revenue share.

Why This Works

No free tier friction: CSP entities pay CHF 25K+ to banks. CHF 99/mo is a rounding error — self-qualifies serious businesses
Aligned incentives: CSP earns more when entities are active across SMB + Enterprise — motivated to push adoption
Four products, one platform: CSP entities get co-branded SMB + Enterprise. Direct merchants get SMB. All get Private Banking. Same infrastructure.
B2B2B flywheel: Enterprise clients' payees become SMB merchants. CSP entities use Enterprise for their own payouts. Each product feeds the others.
CSP as distribution: Each CSP brings 20-100 entities at zero CAC. Every entity is a customer for all four product lines from day one.

Revenue Projection (Year 1 — All Products, Switzerland)

Product Line	Revenue	% of Total
Kasha SMB (direct) — 300-500 merchants	~CHF 470K	56%
Kasha Enterprise — 5-15 clients	~CHF 100K	12%
CSP entity revenue — 80-150 entities (SMB + Enterprise)	~CHF 185K	22%
CSP platform + formation fees	~CHF 70K	8%
Private Banking commissions	~CHF 40K	5%
Revenue share paid to CSPs	~CHF -22K	—
Total Year 1 (Switzerland)	~CHF 843K	100%

Implementation: Provisioning API Mapping

CSP-level plans: New CSP_STARTER, CSP_PROFESSIONAL, CSP_ENTERPRISE in ServiceKey enum
Formation fees: Existing COMPANY_FORMATION ServiceKey, triggered via processCompanyFormationPayment
Entity provisioning: CSP entities get SMB_PRO / SMB_BUSINESS / SMB_PREMIUM tiers with referredByCSP flag. No free tier — minimum SMB_PRO enforced. Optional ENTERPRISE add-on.
Product line routing: productLines array on tenant (smb, enterprise, private_banking) controls which features are exposed in lr-web-app UI
Revenue share engine: Track all entity activity (SMB + Enterprise + Private Banking) by CSP source. Monthly payout calculation via webhooks in lr-banking-service
New build: Revenue share tracker, CSP payout calculator, CSP dashboard in lr-admin-ui, product line router, Enterprise payout module, Private Banking referral tracker

Kasha Enterprise — Voucherised Payments

One-to-many disbursements with near-zero barriers. No KYC from sender. Built on Mass Payout v2 engine. B2B2B feeder for SMB — every payee is a potential SMB merchant.

Fee Type	Amount	Notes
Enterprise Subscription	CHF 499-2,500/mo	Platform access, dashboard, API, compliance tools. Tiered by volume.
Per-Payout Fee	0.5-1.5% of batch	Decreases with volume. Minimum CHF 0.50 per payout.
Voucher Breakage	~3-5% of issued	Unclaimed after 1-year lifecycle. Revenue recognition at expiry.
Cross-Border Fee	0.5-1.0% additional	Multi-currency conversion + compliance.

B2B2B Conversion

Every payout is an SMB lead. Enterprise client disburses to 50 suppliers. Each supplier claims payout, sees the Kasha ecosystem. "You just got paid — want to accept TWINT / Pix too?" At 3-5% conversion, 5-15 Enterprise clients generating payouts to 500+ suppliers yields 25-75 new SMB merchants at zero CAC.

Kasha Private Banking

Not a bank. Licensed agents reselling at least 2 bank products directly. Available to all Kasha customers — SMB merchants, CSP entities, Enterprise clients. Commission-based revenue model.

Revenue Stream	Model	Notes
Account opening commission	Per account	Business account opening via partner banks
Product referral fees	Ongoing commission	Revenue share from banking partner on referred products
Cross-sell uplift	LTV increase	Banking users have higher retention and LTV on core Kasha products

Strategic value: Private Banking isn't just revenue — it's lock-in. Once a merchant's banking is integrated with Kasha, switching cost rises dramatically. The banking relationship deepens engagement with SMB, Enterprise, and CSP products. Year 1 estimate: ~CHF 30-50K in commissions, growing with user base.

Marketplace Pricing (Proposed)

Marketplace access included in SaaS subscription — no extra fee. Commission is seller-side. Research based on Amazon, Mercado Livre, Shopify, Square, Faire, Magazine Luiza, Galaxus benchmarks.

Transaction Commission (on Kasha marketplace)

Monthly Merchant GMV	UK (£)	Switzerland (CHF)	Brazil (BRL)	Rate
Tier 1	First £5,000	First CHF 5,000	First BRL 25,000	8%
Tier 2	£5,001-25,000	CHF 5,001-25,000	BRL 25,001-125,000	6%
Tier 3	£25,001-100,000	CHF 25,001-100,000	BRL 125,001-500,000	4.5%
Tier 4	Above £100,000	Above CHF 100,000	Above BRL 500,000	3.5%

Payment processing passed through at cost (Adyen ~1.2-1.9%, PIX ~0%). Not marked up.

Why 8% starting rate? Below Amazon (8-15%), Mercado Livre (10-19%), Faire (15%), Galaxus (5-25%). Our merchants already pay SaaS fees — combined cost must feel fair. Tiers reward growth and create retention.

Channel Syndication (via lr-channel-manager-service)

Channel	Monthly Fee (UK/CH/BR)	Commission
Amazon SP-API	£29 / CHF 35 / BRL 99	2% of order value
Google Shopping	£19 / CHF 25 / BRL 69	0% (Google is free)
eBay / Etsy	£19 / CHF 25 / BRL 69	2% of order value
Shopify sync	£19 / CHF 25 / BRL 69	1.5% of order value

Revenue Projection (Year 1, UK)

Metric	Value
Active merchants on marketplace	200 (of existing SaaS base)
Average monthly GMV per merchant	£8,000
Blended commission rate	~7.2%
Annual commission revenue	~£1.38M
Channel syndication (50 merchants)	~£17K

Implementation

Checkout flow: Buyer sees product + shipping + tax. Commission is seller-side, not visible to buyer.
Settlement: Adyen for Platforms split-payment. Kasha retains commission + processing, disburses remainder to merchant's Kasha banking account.
Commission invoicing: Auto-generated weekly/bi-weekly via lr-invoicing-service (required for VAT in all 3 markets).
New schema: commission_config (tier rules per merchant), commission_ledger (per-transaction), merchant_payout (settlement cycle).
Revenue recognition: Commission recognized at fulfillment + return window close (IFRS 15). Processing is net revenue (agent, not principal).

Heidi (KashaCH) — TLC Workspace Plan

KashaCH (codename Heidi) is the CSP entity management platform. Full-stack monorepo: Vite + Express + Drizzle + PostgreSQL. TLC test-led development methodology. Deployed via Docker on VPS (161.35.83.210).

Property	Value
TLC Workspace	`kashach-heidi` (alias: `heidi`)
Repo	KashaTech/KashaCH-MonoProject (monorepo)
Stack	React 19, Vite 7, TailwindCSS 4, Radix UI, Express.js, Drizzle ORM, PostgreSQL
AI Agents	Claude (primary, 70%), Codex (secondary, 15%), Gemini (creative, 15%)
Test Framework	Vitest + Testing Library + jsdom
Test Coverage	226 test files, 5,697 tests (all passing), ~75% coverage
Methodology	Test-Led Development (Red → Green → Refactor)
Dev Server	Docker Compose, port 5001 (internal 5000)
Schemas	54 Zod schema files in `shared/schemas/`
External Services	SendGrid (email), AWS S3/GCS (storage), SumSub (KYC), Anthropic SDK (AI)

Heidi Modules (72)

All modules live under server/modules/. Each contains controllers, services, and routes. Schemas in shared/schemas/.

Domain	Modules	Description
Entity Management	`company`, `company-service`, `formation-config`, `formation-document`, `formation-pdf`, `formation-rules`, `formation-email`, `formation-bottleneck`, `formation-completion`, `board-member`, `shareholder`, `person`	Core entity lifecycle: formation (AG, GmbH, Foundation, Ltd, Plc), directors, shareholders, persons, documents, bottleneck tracking
Compliance & Obligations	`compliance`, `obligation`, `obligation-rules-engine`, `obligation-scheduler`, `obligation-task-automation`, `recurring-obligation-scheduler`, `vat-compliance`, `deadline-checker`	Automated obligation tracking, recurring schedules, VAT compliance, deadline monitoring
KYC / KYB	`kyc`, `aml`	SumSub integration, AML checks, beneficial ownership verification
Notarization	`notarization-cases`	Blue ink document management, QR codes, notary case tracking
CSP Management	`csp`, `platform`, `platform-settings`, `jurisdiction`, `onboarding-template`, `onboarding-automation`	Multi-tenant CSP configuration, white-label settings, jurisdiction rules, onboarding workflows
User & Access	`auth`, `user`, `user-settings`, `roles`, `rbac`, `membership`, `invitation`	Authentication, role-based access control, team invitations, user preferences
Tasks & Automation	`task`, `task-template`, `automation`, `automation-trigger-engine`, `activity`, `timeline`	Task management, templates, automation rules, activity feeds, entity timelines
Billing & Invoicing	`billing`, `catalog`	Service catalog, billing calculations, invoicing
Documents & PDF	`document`, `pdf`, `email-template`	Document storage, PDF generation (pdfmake), email templates
Leads & Partners	`leads`, `lead-activity`, `lead-forms`, `partners`, `partner-activity`, `partner-notification`, `partner-portal`, `pipeline-stages`	Lead capture, CRM pipeline, partner portal, UTM tracking, commission management
External Portals	`external-professionals`, `public`, `landing-page-templates`	External portal for notaries/auditors/accountants, public formation wizard, CSP landing pages
Communication	`notification`, `notification-routing`, `shipment-email`, `shipping`	Notification system, routing rules, shipment tracking emails
Search & UI	`search`, `saved-view`, `column-preferences`, `tag`, `metrics`, `admin`	Full-text search, saved views, column customization, tagging, metrics dashboards
AI	`server/agent/` (Heidi AI)	Context-aware AI assistant using Anthropic SDK — entity context, compliance guidance, formation help
Integration	`integration`, `uid`, `zefix`	External integrations, Swiss UID registry, Zefix company register

Client Pages

Section	Pages
Core	Dashboard, Companies, People, Formations, Compliance, Tasks, Notifications
Commercial	Billing, Leads, Partners, Partner Portal
Admin	Admin, Settings, User Settings, Onboarding
External	External Portal, Public Wizard, Marketing, Login, Invite, Privacy, Terms
Shipping	Shipments (notarized document delivery tracking)

Heidi Phase Roadmap

32 phases defined. TLC test-led methodology on all. Milestone: Go Live.

Completed Phases

#	Phase	Highlights
01	Platform Independence	Removed Replit dependency, Docker deployment
02	Code Consolidation	56 modules migrated, utilities consolidated, reusable component library
03	Company Formation Wizard	AG, GmbH, Foundation formation flows
04	Notary Integration	Blue ink docs, QR codes, notary case tracking
05	Formation-Onboarding Integration	Connected formation → onboarding pipeline
11	UI Tooltips	Contextual help throughout the UI
12	Contextual Heidi	Context-aware AI assistant (Anthropic SDK)
22	Leads Module	Bug fixes, service extraction, tests
26	Remove Hardcoding	All hardcoded values externalized to config
27	External Portal Enhancements	File upload, notifications, multi-role support
29	Deduplicate Components	Shared components, billing calculations, service-layer enforcement

In-Progress & Urgent

#	Phase	Status	Owner	Description
28	Company Formation Improvements	In Progress	@jurgen	UAT fixes, quote flow, CSP config, capital deposits — 19 sub-tasks with dedicated test plans
29b	Pre-Launch Schema Audit	CRITICAL	Claude + Codex	DB hardening, schema redundancy audit, data dictionary, migrations review
30	External Portal Role Views	In Progress	@divya	Role-differentiated views for Notary, Auditor, Accountant portals
23	Partners Module	In Progress	@divya	Landing pages, UTM tracking, partner portal, commission management

Planned (Go-Live Path)

#	Phase	Priority	Description
06	CSP Lead Generation	Tier 5	Calendar integration, lead capture pages from public wizard
07	E2E Testing	Tier 2	Playwright end-to-end tests — requires stable features
08	Wiki Documentation	Tier 2	API docs, Swagger, knowledge base
09	User Manuals	Tier 3	Screenshot-based guides (depends on stable UI)
10	Marketing Website	Tier 5	Public-facing site for CSP acquisition
24	KYB & Jurisdiction Compliance	Planned	KYB verification, jurisdiction-specific rules engine
25	Tax Rules for Invoicing	Planned	VAT/GST calculation, jurisdiction-based tax
31	Heidi Agent Improvements	Tier 2	Streaming, native tool calling, rate limiting, token tracking, knowledge optimization
32	TypeScript Strict Mode	Planned	Fix 86 pre-existing type errors, re-enable `npm run check` in CI

Future (Post Go-Live)

#	Phase	Description
13	Architecture Migration	Move to Next.js + Nest.js
14	Operations & Observability	Monitoring, backups, alerting
15	Security & Compliance	Pen testing, GDPR, SOC 2
16	Localization	de/fr/it translations
17	Mobile & PWA	Offline support, push notifications
18	Customer Onboarding	Product tours, support flows
19	Public API & Integrations	Webhooks, 3rd party apps
20	Analytics & Reporting	Dashboards, custom reports
21	Performance & Scaling	Caching, horizontal scaling

Phase Dependencies

    01 Platform Independence → DONE

    02 Code Consolidation → DONE

    03 Formation Wizard → 05 Formation-Onboarding → 04 Notary → DONE

        ↓

    28 Formation Improvements (in progress)

        ↓

    07 E2E Testing ← needs stable features

        ↓

    08 Wiki Docs → 09 User Manuals (needs stable UI for screenshots)

    22 Leads → 23 Partners (in progress) → 06 Lead Gen → 10 Marketing Website

    29b Schema Audit (CRITICAL) → blocks Go-Live

    30 External Portal Role Views (in progress) → blocks Go-Live

Team & Quality

Member	Role	Current Focus
Jurgen Calleja	Developer	Phase 28 (Formation Improvements), Phase 31 (Heidi Agent)
Divya	Developer	Phase 23 (Partners), Phase 30 (External Portal Roles)
Blair Avellino	QA	UAT, test coverage

Quality Metrics

Metric	Current	Target
Test files	226	250+
Test cases	5,697 (all passing)	6,500+
Coverage	~75%	80% (TLC threshold)
TypeScript errors	0 (86 in strict mode)	0 in strict mode
Quality score threshold	75	80+
Shared schemas	54 Zod files	All validated at runtime

TLC LLM Router

Multi-model capability routing configured at workspace level:

Capability	Primary	Fallback	Mode
Code review	Claude + Codex	Claude	Consensus
Code generation	Claude	Codex	Single
Refactoring	Claude	—	Single
Design / Vision	Gemini	Claude	Single
Brainstorming	Gemini + Claude	Claude	Parallel

Bucket List

Ideas beyond the current sprint cycle. High-value, not yet scoped.

Idea	Market	Notes
Anticipation on competitor processing	Brazil	File claims via the national settlement registry (Registradora de Recebíveis / CIP) — the Central Bank's infrastructure for receivables registration. Even if a merchant processes with a competitor, the receivable is registered federally and we can place a claim for anticipation. This is the foundation of antecipação at the federal level.
Multi-acquirer anticipation engine	Brazil	Aggregate receivables across all acquirers (Cielo, Rede, Stone, PagSeguro) via CIP registry for a unified credit line
Anticipation on recurring B2B payouts	Global	Extend the Carrefour/anchor-buyer model to any B2B relationship with predictable payment cycles
Jurisdiction #3+	EU / Dubai	Heidi CSP expansion beyond CH + UK

Kasha & Heidi Project Plan · Q1-Q2 2026 · Generated by Claude & Gemini · Authored by JJ