Hi Product Hunt 👋 I'm a solo dev and Lacis is the tool I wished existed. Every new AI chat forgets who I am and how I work. So I keep re-pasting the same context — my stack, my rules, my conventions. Lacis fixes that: you set your context *once*, as a small tree of typed branches (identity, rules, architecture, skills), and a router auto-injects only the relevant pieces into every reply — with a banner showing you which ones loaded. It's like a living, selective CLAUDE.md / .cursorrules, reusable across projects. Then you branch. The whole conversation is a tree: branch from any message, and each branch only ever sees its own path — never its siblings. So you can explore several approaches in parallel without the model mixing them up, and your context tree rides along on every branch. It lives in VS Code: • Uses the models in your existing GitHub Copilot plan — no extra key, no second bill. Or bring your own OpenAI / Anthropic / OpenAI-compatible key. • Picks up your MCP tools, and can read/write/edit files with native diff approval on every change. • Local by default — no account, no cloud. Share a tree as a read-only link. ▶️ 90-second demo: I build a Mario-style platformer with only AI prompts — set the context tree, then branch two jump feels and two levels side by side: https://youtu.be/t2uZLrNtNF4 Free on the VS Code Marketplace (and Open VSX for Cursor / Windsurf / VSCodium). It's early (v0.0.4) and closed source for now — I'd love your honest take on whether the context-tree idea clicks.

Lacis

Set your AI's context once — it follows every branch

Set your AI's context once — it follows every branch