Unclassified

IntegrateWise — Architecture Context & Documentation Rules

Master reference for all agents, contributors, and documentation authors.
Last updated: April 2026

Product Definition & Terminology
Architecture Overview
The Three Data Flows
L1 — The Workspace (12 Domain Views)
L2 — The Cognitive Layer (14 Components)
The Schema System
Entity 360 — Convergence Point
The Cognitive Loop
Connector Ecosystem
The Knowledge Memory
Trust & Governance
20 Products Across 3 Surfaces
Founding Story
Go-To-Market (Current Phase)
Architectural Review (9.2/10)
Documentation Rules — Public vs Internal
Writing Style
Sidebar Structure — Public Docs
Known Issues (Code, Not Docs)
Key Source Files

1. Product Definition & Terminology

What IntegrateWise IS

IntegrateWise is a Knowledge Workspace over the Spine, empowered by AI.

It is NOT a platform, NOT an OS, NOT a tool.

Approved Terms

Term	Usage	Context
Knowledge Workspace	Primary descriptor	Always use this
Spine	Default name for the intelligent data layer	Product-level term
Adaptive Spine	Variant of Spine	Glossary synonym — same thing
Intelligent Spine	Variant of Spine	Glossary synonym — same thing
Unified Intelligence Layer	Variant of Spine	Glossary synonym — same thing
Adaptive Layer	Variant of Spine	Glossary synonym — same thing
Cognitive Intelligence Overlay	L2 overlay experience	Correct as-is — "overlay" is intentional
Cognitive Layer	Short form of above	Acceptable
TruthLayer	Trust/governance system	Platform capability, not a product
Entity 360	Convergence point for all 3 flows	Core view, not standalone product

NEVER Say Externally

"We are a platform"
"We are an AI tool"
"Cognitive OS"
"IntegrateWise OS"
"We replace Gainsight"
"We are like Salesforce but with AI"

Hero Anchor

"Your tools don't talk to each other. Your AI doesn't know context. You are the bridge. The Human API."

Trust Line

"Every AI remembers. Only IntegrateWise remembers the truth."

2. Architecture Overview

IntegrateWise has two product layers:

L1 — The Workspace

12 domain-specific views
Schema-filtered per tenant (industry + department)
What users see and work in daily

L2 — The Cognitive Layer (Overlay)

14 components accessible via Cmd+J
Non-intrusive overlay — sits alongside work, extends when needed
Also schema-filtered

Platform Engine (Under the Hood)

The Spine (intelligent plumbing connecting all data)
Entity 360 (convergence point)
Schema system (12 domains × 11 industry overlays)
8-stage processing pipeline (INTERNAL ONLY — never expose publicly)
Connector ecosystem
Knowledge Memory

┌──────────────────────────────────────────────────┐
│                   L2 — COGNITIVE OVERLAY          │
│  (SpineUI, ContextUI, KnowledgeUI, Twin, HITL,   │
│   Signals, Think, Act, Govern, Adjust, Repeat,    │
│   Audit Trail, Agent Config, Evidence Drawer)      │
├──────────────────────────────────────────────────┤
│                   L1 — WORKSPACE                  │
│  (12 domain views, schema-filtered per tenant)    │
├──────────────────────────────────────────────────┤
│              PLATFORM ENGINE                      │
│  Spine │ Entity 360 │ Schema │ Pipeline │ Memory  │
├──────────────────────────────────────────────────┤
│              CONNECTORS                           │
│  50+ SaaS │ MCP │ Webhooks │ Accelerators │ AI    │
└──────────────────────────────────────────────────┘

3. The Three Data Flows

Three distinct data flows with hard boundaries:

Flow A — Structured / Truth

Tools → Loader → Normalizer → Spine
                                 ↓
                         Actions on tools
                                 ↓
                         (flows back into Flow A — REPEAT LOOP)

Source: SaaS connectors (CRM, billing, support, etc.)
Destination: The Spine (structured truth)
Has a repeat loop: governed actions executed on tools flow back as new structured data
This is the ONLY flow with a repeat loop

Flow B — Unstructured / Context

Docs, emails, conversations → Context Store + Spine linking

Source: Documents, emails, Slack messages, meeting transcripts
Destination: Context Store, linked to Spine entities
NO repeat loop
Enriches entities with qualitative context

Flow C — AI Knowledge / Memory

AI sessions → Triage Bot (SOLE writer) → Memory Store

Source: AI interactions, Twin sessions, user conversations
Writer: Triage Bot is the SOLE writer to memory
Twin can also write Truth + Context summaries via Triage Bot
Read access: per tenant OAuth only
Grows daily — accumulated intelligence
NO repeat loop
NEVER writes directly to the Spine — only via HITL approval

Critical Boundaries

Property	Flow A	Flow B	Flow C
Data type	Structured	Unstructured	AI-generated
Destination	Spine	Context Store	Memory Store
Repeat loop	YES	NO	NO
Writes to Spine	Direct	Linked	NEVER (only via HITL)
Trust level	Verified	Contextual	Requires approval

4. L1 — The Workspace (12 Domain Views)

Each domain view is schema-filtered and provides role-specific interfaces.

#	Domain	Icon	Files	Views	Entity Types	Notes
1	Account Success	💚	49	20 active	18	Most feature-rich domain
2	BizOps	⚙️	27	7 + C-suite views	14	CEO/COO/CIO/Founder views
3	Finance	💰	20	7	15	Revenue, billing, forecasting
4	IT Admin	🔧	31	6	12	System management
5	Marketing	📢	16	8	14	Campaigns, attribution
6	Personal	👤	10	8	7	Waitlisted for launch (M1-6)
7	Procurement	📦	11	6	11	Vendor management
8	Product Engineering	💻	14	7	16	Roadmap, sprints, releases
9	RevOps	📈	8	8	14	Most analytically sophisticated
10	SalesOps	🎯	5	8	14	Pipeline, forecasting
11	Service	🎧	11	6	12	Support, SLA, escalation
12	Student-Teacher	🎓	—	—	—	Shell only (0 bytes, dashboard wired, not in DomainId)

Source Locations

Domain implementations: apps/web/src/components/l1/domains/
Domain type configs: apps/web/src/components/l1/domains/domain-types.ts
Schema definitions: packages/types/src/schema.ts (DOMAIN_SPINE_CONFIG)

5. L2 — The Cognitive Layer (14 Components)

Accessible via Cmd+J. Two overlay implementations exist in the codebase:

#	Component	Purpose
1	SpineUI	View and interact with Spine data
2	ContextUI	Browse unstructured context
3	KnowledgeUI	Access knowledge memory
4	Evidence Drawer	Supporting evidence for signals
5	Signals	AI-generated insights (from Think layer)
6	Think (Active Situations)	Situation awareness, pattern detection
7	Act	Agent execution (Twin lives here)
8	HITL / Approvals	Human-in-the-loop approval queue
9	Govern	Governance rules and boundaries
10	Adjust	Configuration and tuning
11	Repeat	Loop monitoring and status
12	Audit Trail (LayerAudit)	Full audit log of all actions
13	Agent Config	Configure agent behaviors
14	Digital Twin	Personal AI agent with read-only Spine access

Important Notes

Signals are AI Insights generated in the Think layer — NOT raw events from connectors
Agents (including Twin) live ONLY in Act, not in truth formation
Twin has READ-ONLY access to Spine; can only write via HITL approval
Two overlay implementations exist: intelligence-overlay-new.tsx (bottom-sheet) and cognitive/CognitiveLayer.tsx (sliding panel, 13 tabs)

6. The Schema System

The schema system is IntegrateWise's real moat — runtime composability per tenant.

How It Works

getSpineConfig(industry, department) → unique schema per tenant

12 domain schemas × 11 industry overlays = combinatorial configuration
Schema controls EVERYTHING downstream: loader extraction, normalizer filtering, L1 views, L2 cognitive scope

11 Industry Overlays

SaaS, Healthcare, Manufacturing, Automotive, Retail, Financial Services, Logistics, Media & Entertainment, Public Sector, Education, Professional Services

Authoritative Sources

File	Lines	Content
`packages/types/src/schema.ts`	2113	`DOMAIN_SPINE_CONFIG` — 12 domains (authoritative)
`packages/types/src/adaptive-schema.ts`	—	10 domains (DRIFT — missing 2)
`sql-migrations/050_adaptive_spine_12_domains.sql`	1130	Database schema for all 12 domains

Known Drift

adaptive-schema.ts only defines 10 of 12 domains (missing Procurement and Student-Teacher)
domain-types.ts has stale spineProjection (only "bizops"/"sales" but should be 12 values)

7. Entity 360 — Convergence Point

Entity 360 is where all three data flows converge:

Flow A (Structured Truth)     ──┐
Flow B (Unstructured Context) ──┼──→ Entity 360
Flow C (AI Knowledge/Memory)  ──┘
                                       ↓
                                 Twin activates here
                                 (READ-ONLY access to Spine)

Entity 360 requires cross-source data to be meaningful
You CANNOT "connect one system and see Entity 360" — this was a doc error
Twin can only write to Spine via HITL approval from this context

8. The Cognitive Loop

Think → Act → Govern → Repeat → Adjust
  ↑                                 │
  └─────────────────────────────────┘

Stage	What Happens
Think	Pattern detection, signal generation, situation awareness
Act	Agent execution — Twin and other agents live here
Govern	Rule enforcement, trust boundaries, compliance checks
Repeat	Loop execution on structured data (Flow A only)
Adjust	System tuning, confidence calibration, feedback loops

Critical: Where Agents Live

Agents (including Twin) live ONLY in Act
Agents do NOT participate in truth formation
All agent outputs go through Govern before affecting the Spine

9. Connector Ecosystem

Connector Types

Type	Description	Examples
SaaS Connectors	Standard business tool integrations	Salesforce, HubSpot, Stripe, Zendesk
MCP Connectors	Model Context Protocol integrations	Knowledge base, Figma
Webhooks	Event-driven integrations	Custom event sources
Accelerators	Pre-built domain packages	Industry-specific starter configs
AI Connectors	AI service integrations	ChatGPT, Claude, Gemini, Perplexity, Grok

Key Principles

"Every input is a connector" — unified ingestion model
Auto-upgrade system: child connectors inherit upgrades from parent connectors
Purchase decision entry point: users check connector availability first
Only claim production-verified count: 4-point verification (OAuth flow, creamy load, field mapping, delta sync)
Connector count at launch: state only what passes verification (DO NOT claim "50+")

Credential Storage

Tokens ARE stored encrypted in connector_credentials
Previous docs said "No credential storage" — this was FALSE and must be corrected
Correct framing: "Credentials are encrypted at rest, scoped per tenant, with OAuth refresh rotation"

10. The Knowledge Memory

This is what makes IntegrateWise a "Knowledge Workspace":

Accumulated intelligence that grows daily
No retraining needed — context is always current
Triage Bot is the sole writer to memory
Per-tenant OAuth scoping for read access
Twin reads verified memory, writes back summaries via Triage Bot

Why It Matters

Traditional AI tools lose context between sessions. IntegrateWise's Knowledge Memory:

Remembers every verified interaction
Builds organizational knowledge over time
Provides the Twin with historical context
Never requires manual knowledge base maintenance

11. Trust & Governance

The TruthLayer

TruthLayer is a platform capability, not a standalone product.

How it works:

AI generates output (entity extraction, recommendations, summaries)
Triage Bot processes: confidence scoring, conflict detection
Results go to Human Approval queue
Approved items become verified memory
Twin reads verified memory at Entity 360

HITL (Human-in-the-Loop)

Every AI action that affects the Spine requires human approval
Three levels: auto-approve (high confidence), review (medium), block (low/conflict)
Approval queue visible in L2 Cognitive Layer

Trust Boundary

Flow A and B write directly to their stores (Spine, Context Store)
Flow C NEVER writes directly to Spine
The boundary between AI-generated and verified data is absolute

12. 20 Products Across 3 Surfaces

Surface 1: Account Success (8 Products)

No Twin:

DataSentinel — Data quality monitoring
VaultGuard — Compliance and audit
ArchitectIQ — Solution architecture
TemplateForge — Template management

Basic Twin:

SuccessPilot — CS automation
DealDesk — Deal management

Full Twin + TruthLayer:

ChurnShield — Churn prediction and prevention
SuccessCommand — Executive CS dashboard

Surface 2: Business Ops (7 Products)

No Twin:

ComplianceVault — Regulatory compliance
VendorGuard — Vendor management
PartnerBridge — Partner ecosystem

Basic Twin:

GrowthDesk — Growth analytics
HirePilot — Hiring automation

Full Twin + TruthLayer:

FinPulse — Financial intelligence
OpsCore — Operations command center

Surface 3: Personal Space (5 Products)

WAITLISTED for M1-6 launch. CTA: "Join the Waitlist"

No Twin:

WealthPilot — Personal finance
WellnessCore — Health and wellness

Basic Twin:

LearningDesk — Learning management
RelationshipMap — Professional network

Full Twin + TruthLayer:

LifeOps — Personal operations hub

13. Founding Story

Nirmal was a CSM (Customer Success Manager)
Experienced the Human API problem firsthand — being the bridge between disconnected tools
Built the Spine (intelligent plumbing) which saved an $8M account
Multiple other proofs: support cases solved, unresolved issues resolved
Proposed organization-wide implementation
Superiors tried to grab the idea
Realized the opportunity was bigger — started exploring independently
Started as Mule Nexus in 2024, evolved into IntegrateWise
"Depth matters" — Depth Matrix stays as a concept

14. Go-To-Market (Current Phase)

M1-3: Founder-Led Sales

Element	Value
CTA	"Book a Demo" / "Join Early Access"
Model	Founder-led sales (NOT self-serve)
Personal Space	Waitlisted (not live)
Pricing tiers	Starter, Growth, Command

What NOT to Say on the Website

"Start Free" / "Start Free Trial" (not until M3-6)
"No credit card required"
"Connect in 2 minutes"
"50+ integrations" (only claim verified count)
Self-serve signup flow references

CTA Progression

Phase	CTA	Model
M1-3	Book a Demo	Founder-led
M3-6	Start Free Trial	Guided PLG
M6+	Start Free	Full PLG

15. Architectural Review (9.2/10)

Strengths

Schema-driven architecture — real moat, runtime composability per tenant
Three-flow separation with hard boundaries — clean data trust model
Trust/governance model — Triage Bot + HITL + tenant OAuth
Continuous loop via Flow A — the only flow with repeat
Knowledge memory accumulation — commercial moat that grows daily

Issues (Code-Level, Not Documentation)

Issue	Location	Impact
Stale `spineProjection`	`domain-types.ts`	Only "bizops"/"sales" but should be 12 values
Schema duplication	`schema.ts` vs `adaptive-schema.ts`	12 vs 10 domains
Two competing L2 overlays	`intelligence-overlay-new.tsx` vs `CognitiveLayer.tsx`	UX inconsistency
Account Success disproportionately deep	49 files vs 5-16 for others	Coverage gap
Student-Teacher shell empty	0 bytes, not registered in DomainId	Incomplete domain

16. Documentation Rules — Public vs Internal

What Public Docs SHOULD Cover

The product story (Human API problem → Turning Point → Solution)
The Workspace (what users see, role-based views, schema-driven)
The Spine (intelligent plumbing, what it connects, adaptive, grows with you)
The Cognitive Layer (overlay experience, not internals)
Three Data Flows (at PRODUCT level, not architecture level)
Knowledge Memory (accumulated intelligence, no retraining)
Connectors (purchase decision page + auto-upgrade + types)
Who It's For (founders, CS, sales, finance, product, analysts)
Trust & governance (truth boundary, approval-driven, HITL)

What Public Docs MUST NOT Expose

8-stage pipeline internals
Normalizer stages (S1-S8, S7.5)
Hydration buckets (B0-B7), "creamy load"
Twin trigger engine mechanics (10 triggers, guardrails, scoring)
Identity resolution internals
Worker-level architecture (Cloudflare Workers, D1, KV, queues)
Firebase SMB architecture
Any backend processing details — the backend is infrastructure, NOT the product

Errors in Previous Public Docs (All Fixed in Rewrite)

Error	Correction
Homepage tagline "System of Context"	Use "Knowledge Workspace"
Hero CTA "Get Started"	"Book a Demo"
Getting Started describes self-serve signup	Remove — no self-serve at launch
"Connect first system, see Entity 360"	Entity 360 needs cross-source data
"50+ business systems"	Only claim production-verified count
"No credential storage"	Tokens stored encrypted in `connector_credentials`
"Bi-directional sync"	Governed action execution
Custom Connector SDK referenced	Doesn't exist
Solutions: 5 role pages	Align to "Who It's For" framing
Technical pipeline details in public docs	Move to internal

17. Writing Style

Voice & Tone

Authoritative but approachable
Second-person ("you")
Active voice
Product-first, not tech-first
No emojis in prose

Page Structure

Every page must have:

---
title: Page Title
slug: /section/page-slug
summary: One-sentence summary
audience: [public | internal | team]
status: [draft | review | published]
owner: nirmal
lastReviewed: 2026-04-09
publishEligible: true
tags: [relevant, tags]
---

# Page Title

Opening statement (1-2 sentences).

> Blockquote principle (where appropriate)

## Section Heading

Content...

## Next Steps

- [Link to next page](/section/next-page)
- [Link to related page](/section/related-page)

Formatting Conventions

ASCII diagrams for architecture (not images)
Tables for comparisons
VitePress containers (::: tip, ::: info, ::: warning) used sparingly
Five Core Principles referenced throughout where appropriate

18. Sidebar Structure — Public Docs

The Story (3 pages)
├── Why Work Is Broken         /story/why-work-is-broken
├── The Turning Point          /story/the-turning-point
└── What Is IntegrateWise      /story/what-is-integratewise

The Workspace (4 pages)
├── How It Works               /workspace/how-it-works
├── Domain Views               /workspace/domain-views
├── Entity 360                 /workspace/entity-360
└── Day One Experience         /workspace/day-one

The Spine (5 pages)
├── What the Spine Is          /spine/what-the-spine-is
├── What It Connects           /spine/what-it-connects
├── Adaptive by Design         /spine/adaptive-by-design
├── Grows with You             /spine/grows-with-you
└── The Truth Boundary         /spine/truth-boundary

The Cognitive Layer (6 pages)
├── How the Overlay Works      /cognitive-layer/how-the-overlay-works
├── Spine, Context & Knowledge /cognitive-layer/spine-context-knowledge-ui
├── The Twin                   /cognitive-layer/the-twin
├── Signals & Think            /cognitive-layer/signals-and-think
├── Approval-Driven Execution  /cognitive-layer/approval-driven-execution
└── The Continuous Loop        /cognitive-layer/the-continuous-loop

Three Data Flows (3 pages)
├── Structured Truth           /data-flows/structured-truth
├── Unstructured Context       /data-flows/unstructured-context
└── AI Knowledge & Memory      /data-flows/ai-knowledge-memory

The Knowledge Memory (3 pages)
├── How Memory Works           /knowledge-memory/how-memory-works
├── Why Knowledge Workspace    /knowledge-memory/why-knowledge-workspace
└── No Retraining Needed       /knowledge-memory/no-retraining-needed

Connectors & Integrations (4 pages)
├── Overview                   /connectors/overview
├── How Connectors Work        /connectors/how-connectors-work
├── Auto-Upgrade               /connectors/auto-upgrade
└── Request a Connector        /connectors/request-a-connector

Who It's For (4 pages)
├── Founders & Leaders         /who-its-for/founders-and-leaders
├── Revenue & Customer Teams   /who-its-for/revenue-and-customer-teams
├── Operations & Finance       /who-its-for/operations-and-finance
└── Product & Engineering      /who-its-for/product-and-engineering

Reference (3 pages)
├── Glossary                   /reference/glossary
├── Depth Matrix               /reference/depth-matrix
└── FAQ                        /reference/faq

19. Known Issues (Code, Not Docs)

These are codebase issues identified during architectural review. They do NOT affect documentation content but should be tracked for engineering.

spineProjection stale in domain-types.ts — only "bizops"/"sales" but should cover all 12 domains
Schema duplication — schema.ts (12 domains) vs adaptive-schema.ts (10 domains, missing Procurement + Student-Teacher)
Two competing L2 overlay implementations — intelligence-overlay-new.tsx (bottom-sheet) and cognitive/CognitiveLayer.tsx (sliding panel, 13 tabs)
Account Success disproportionately deep — 49 files, 20 views vs 5-16 files for other domains
Student-Teacher domain shell empty — 0 bytes, dashboard wired but not registered in DomainId enum

20. Key Source Files

Product Content Sources

File	Purpose
`docs/gtm/POSITIONING_FRAMEWORK.md`	Approved positioning, always/never say
`docs/gtm/LAUNCH_PHASE_ADDENDUM.md`	Launch overrides, CTA rules
`docs/product/PRODUCT_CATALOG_20.md`	20 products across 3 surfaces
`docs/product/PRODUCT_LAYERS_BY_ROLE.md`	Role-based views
`docs/product/PLATFORM_FEATURES_COMPLETE.md`	Full feature inventory

Architecture Sources

File	Purpose
`packages/types/src/schema.ts`	12 domain schemas (AUTHORITATIVE)
`packages/types/src/adaptive-schema.ts`	Adaptive schema (DRIFT — 10 domains)
`apps/web/src/components/l1/domains/`	All 12 domain view implementations
`apps/web/src/components/l2/`	All 14 L2 cognitive components
`apps/web/src/components/l1/domains/domain-types.ts`	Domain configs
`sql-migrations/050_adaptive_spine_12_domains.sql`	Database schema

VitePress Config

File	Purpose
`docs/.vitepress/config.ts`	Main VitePress configuration
`docs/internal/_sidebar.ts`	Internal docs sidebar
`scripts/sync-docs.sh`	Sync from live repo to docs

End of Architecture Context. This document is the single source of truth for all documentation decisions.

Table of Contents

1. Product Definition & Terminology

What IntegrateWise IS

Approved Terms

NEVER Say Externally

Hero Anchor

Trust Line

2. Architecture Overview

L1 — The Workspace

L2 — The Cognitive Layer (Overlay)

Platform Engine (Under the Hood)

3. The Three Data Flows

Flow A — Structured / Truth

Flow B — Unstructured / Context

Flow C — AI Knowledge / Memory

Critical Boundaries

4. L1 — The Workspace (12 Domain Views)

Source Locations

5. L2 — The Cognitive Layer (14 Components)

Important Notes

6. The Schema System

How It Works

11 Industry Overlays

Authoritative Sources

Known Drift

7. Entity 360 — Convergence Point

8. The Cognitive Loop

Critical: Where Agents Live

9. Connector Ecosystem

Connector Types

Key Principles

Credential Storage

10. The Knowledge Memory

Why It Matters

11. Trust & Governance

The TruthLayer

HITL (Human-in-the-Loop)

Trust Boundary

12. 20 Products Across 3 Surfaces

Surface 1: Account Success (8 Products)

Surface 2: Business Ops (7 Products)

Surface 3: Personal Space (5 Products)

13. Founding Story

14. Go-To-Market (Current Phase)

M1-3: Founder-Led Sales

What NOT to Say on the Website

CTA Progression

15. Architectural Review (9.2/10)

Strengths

Issues (Code-Level, Not Documentation)

16. Documentation Rules — Public vs Internal

What Public Docs SHOULD Cover

What Public Docs MUST NOT Expose

Errors in Previous Public Docs (All Fixed in Rewrite)

17. Writing Style

Voice & Tone

Page Structure

Formatting Conventions

18. Sidebar Structure — Public Docs

19. Known Issues (Code, Not Docs)

20. Key Source Files

Product Content Sources

Architecture Sources

VitePress Config