cwlin0131/Vault-for-Founders

What it solves

Founders are increasingly handing important decisions to AI — product strategy, market analysis, hiring calls. Every new chat with an AI starts from zero, because the AI has no memory of who you are, what your company does, or what you decided before. Built-in memory in tools like ChatGPT or Claude is opaque, platform-locked, fragile, and can't scale. This project gives founders a way to build their own persistent, portable, version-controlled knowledge base that any AI can read — so the AI starts every session with full context.

How it works

The framework uses two tools: Obsidian (a local Markdown editor) and Git (a version control system). Your knowledge lives in a plain-text folder called a Vault, containing .md files that both you and an AI agent can read directly — no API integration needed. Git tracks every change, letting you roll back mistakes and sync to GitHub for backup. The README includes a full setup guide, architecture notes, optimization guide, and ready-to-use templates for identity, memory, after-action reviews, and more. You hand the repository link to any AI agent (like Claude Cowork or OpenClaw), and the agent will walk you through building your own Vault.

Who it's for

Founders and startup operators who want to delegate meaningful work to AI without losing their company's context. You don't need a technical background — the guide explains terms like Vault, Git, and Repo in plain language. It's also designed for AI agents reading the repo to assist a human user in setup.

Highlights

• Portable and future-proof: Your Vault is your own files, so you can switch AI tools anytime and keep all your memory.
• Version-controlled: Git lets you roll back changes and recover from mistakes.
• Plain-text, agent-friendly: AI reads local .md files directly, no API costs or integrations.
• Founder-focused structure: Templates for identity, decision style, memory summaries, after-action reviews, and standard operating procedures.
• Includes optimization guide: Learn to index, name, tag, and clean memory for better retrieval.