I open‑sourced a self‑hosted “context layer” so I stop re‑explaining my projects to AI

Over the last year I kept hitting the same pain: every new chat, model, or tool starts fresh. I’d re‑explain the same background, goals, and project details over and over.

So I built Easy AI Context — a small MCP server that gives your AI a personal context tree it can read/write.

What it is:

• A Model Context Protocol (MCP) server
• 4 tools only: list_paths, get, set, delete
• Context is a tree of paths (e.g. projects/easy-ai/context)
• Index‑first design: list the tree, read only what you need

Why I think it matters:

• Self‑hosted: your data stays in your AWS account
• Low cost: API Gateway + Lambda + DynamoDB (pay‑per‑use)
• Durable context: works across chats/tools without re‑explaining

Repo: https://github.com/xiaopai20/easy-ai-context

I’m curious how other people are handling long‑term context.
Would you rather keep a small curated memory tree or a larger auto‑logged memory?