Second brain for business: building a secure RAG knowledge system with Outline, Claude, and MCP

We put Eli5's own second brain on the table: how we make company knowledge readable for language models, and why it runs on a wiki, a numbering convention and one MCP server. Floris Schoenmakers and Kishan Chamman, CTO at Eli5, walk through the full system. A second brain for business is a shared, persistent record of how a company works: decisions, documents, meeting notes, playbooks and past projects, kept in a form an LLM can search and reason over. Once that exists, the model answers questions grounded in the company's own knowledge and can build internal tools on top of it. What we cover: The five-layer blueprint: knowledge sources, a self-hosted Outline wiki, Outline's built-in MCP server with SSO, three entry points (Claude Code, Claude chat, Claude Design) and skills in three homes The data structure that makes retrieval unnecessary: stable numbers on every label, an index at the root of every collection, agent-writable and approval-gated collections The feedback loop that keeps the base honest, and the human approval on every write Live demos: company facts, tender and proposal assessment, brand rules, and Figma MCP turning an approved Outline draft into a branded proposal How we limit hallucinations with gates and multi-pass checks Full write-up with both diagrams: [article URL] Modernization assessment: [assessment URL] Chapters 00:00 Intro: the second brain for business 00:42 What a second brain means 01:55 Why it matters now 02:36 System blueprint overview 03:07 Layer 1: knowledge sources 03:39 Layer 2: self-hosted Outline wiki 04:20 Layer 3: MCP access and entry points 05:09 Skills and the feedback loop 07:17 Human in the loop 08:43 Data structure and collections 09:05 Why we run no vector database 11:02 Model choice and vendor lock-in 12:30 Outline versus Notion 14:09 Claude Design and brand adherence 15:00 You need to do the legwork 16:44 Demo: company facts 18:44 Demo: proposal workflow 21:05 Demo: brand rules skill 22:48 Tender filtering in practice 25:43 Figma MCP for proposal styling 29:44 Reducing hallucinations 31:17 Wrap-up Eli5 modernizes legacy platforms and implements AI in process and product. https://www.eli5.io Kishan on LinkedIn:   / kishanchamman   Floris on LinkedIn:   / florisschoenmakers