Wattl AI

The challenge

Business knowledge was scattered across chats and documents. Connected AI tools needed a consistent source of confirmed facts, with people in control of what became trusted knowledge.

The work

Wattl AI is the platform I founded and architect end to end. It began in 2025 as an AI platform for app businesses, generating websites and growth plans through agentic workflows. In 2026 I refocused it on the problem underneath: business knowledge scattered across chats and documents. The product is now the Root, a versioned knowledge base per project, with document import, fact extraction into a review inbox, chat with cited answers, and an MCP server of 30 tools that lets connected AI tools read a context pack for what the user is trying to do and write durable decisions back.

My contribution

Founded the product, narrowed its scope and architected it end to end: React console, Azure API, Python agents, document ingestion, approval workflows, billing and MCP access.

Delivery detail

• Explain the business once; every connected AI tool remembers. Context packs carry confirmed knowledge, approved documents, recent decisions, gaps and conflicts.
• Human control by design: anything an AI tool proposes becomes knowledge only when the owner approves it, and rejected values are remembered and never proposed again.
• Full-stack delivery: Vite and React console, Azure Functions API with Prisma, a Python FastAPI and LangGraph agents service for chat, ingest and indexing, Stripe billing and OAuth-scoped MCP access.
• Shipped and operated: projects, folder and Obsidian-vault import, text-based PDF and ChatGPT-export ingestion, review inbox and history, API keys, audit trail and a 14-day trial.
• Product leadership: the decision to narrow a broad platform vision to a knowledge-first launch, keeping public promises inside working capabilities.

The outcome

A working knowledge platform with document import, a review inbox, cited answers and an MCP server connecting AI tools to shared business context.