tapestree is not assembled from parts. It is grown from three architectural primitives — a unified record graph, a conversation layer that spans every channel, and a per-tenant AI memory that understands decisions, not just documents. Every module is an expression of these three primitives. Nothing is bolted.
In a traditional ERP, data fragments across tables: customers live in CRM, invoices in finance, approvals in email. tapestree dissolves that separation. Accounts, Opportunities, Decisions, Tasks, Events, and Conversations are all records — typed, linked, and traversable through a single graph.
A Decision is not a note. It is a record with an author, a date, a condition, and links to every downstream document it governed. A Task is not a reminder. It is a record with accountability, history, and a conversation thread. When every entity is a record, the platform can reason over it — and so can the people using it.
This is not a schema decision. It is an epistemological one. What gets structured, gets remembered.
The inbox is not a place in tapestree. It is a dimension — every record carries its own conversation thread, and that thread spans every channel simultaneously. Internal team discussion, inbound WhatsApp from a vendor, an outbound email reply, an AI observation — all on the same timeline, against the same record.
External communications do not need routing rules. Inbound WhatsApp from a known contact lands on their record automatically. Outbound messages pick a channel at send time — Email, WhatsApp, or SMS — without leaving the context of the work.
AI participates as a first-class participant. It does not summarise the thread after the fact. It cites from institutional memory in real time — surface the pattern, name the risk, suggest the action.
Every decision, conversation, clause, and event is indexed in a per-tenant vector graph — ACL-aware, PII-redacted, multilingual, and always citable. tapHistory does not search documents. It understands the reasoning behind them.
"Why did we extend credit terms for Global Infra?"
"ABC Industries ke saath 45-day terms kyun approve kiye?"
"Chennai warehouse-il last audit eppotu nadanthathu?"
A bolted platform acquires capabilities by adding products. Each new tool has its own schema, its own conversation layer, its own concept of a customer or a vendor. The integrations multiply. The data diverges. The institutional memory never coheres.
A woven platform is architecturally different. The record graph exists before any module does. The conversation primitive is not a feature — it is the infrastructure. Every module shares the same identity layer, the same record model, the same AI context. Adding a module does not add complexity. It deepens the weave.
The competitive consequence is asymmetric. A bolted competitor can copy a feature. It cannot copy the structural coherence of a platform grown from a single root.
Every product decision in tapestree traces back to one of four commitments. They are not values. They are constraints — and constraints are where craft begins.
Defaults are decisions. Every field, every flow, every state transition was chosen deliberately. There are no "optional" modules that degrade gracefully — everything is first-class or it is not in the platform.
Software that respects the user's context — their language, their channel preference, their regulatory reality — is hospitable. tapestree speaks in the register of the person using it: Hindi, Tamil, or Malayalam; WhatsApp or Email; desktop or phone.
Compliance is not a localisation flag. GST, DPDPA, MSME §15, IndAS, ONDC, Aadhaar e-sign — and the equivalent constructs as we expand — are part of the structural model, not surface plugins.
AI that surfaces every pattern at maximum volume is noise. tapestree AI surfaces the right signal at the right moment — cited, confidence-scored, and actionable. Precision over volume. The operator decides; the platform informs.
Structure is memory. Memory is leverage.
A navigable mockup of the full platform — every module, every tab, every record open. No login. No sales call required to explore the architecture.