Hey everyone, Like many of you, I find myself jumping between different AI coding environments (Claude Code in the terminal, Gemini CLI, Cursor, Copilot, etc.). The problem is that each tool hovers over its own configs: If you set up a useful database MCP server in Claude Code, Cursor/Gemini knows nothing about it. If you write a custom prompt helper/skill in Gemini CLI, Claude Code starts as a blank slate. You end up manually copy-pasting .md files or duplicating configurations across five different directories. To solve this, I built Wasla (Arabic for "connection") — a lightweight, zero-dependency CLI tool and daemon that syncs your agent instructions and MCP configurations across all your AI assistants. How it works: No Duplication: Wasla scans your configured tools (~/.claude/, ~/.gemini/, etc.) and registers assets. Latest is Greatest: Whichever tool you used most recently to edit an agent/instruction becomes the source of truth for the next sync. Reference Stubs: Instead of copying entire files, Wasla writes lightweight pointers/stubs. The native CLI tool reads the stub, follows the pointer to the original, and loads it instantly. Core Features: Scope Support: Use --scope workspace to keep sync local to a project, or --scope user to share agents globally across your machine. Background Sync (Watch Mode): Runs in the background, detects file saves, and instantly synchronizes changes across all tool directories. Interactive Dashboard: Running "wasla visualizer" opens a React-based visualizer map showing exactly how your agents, instructions, and MCP connections are mapped across your machine. It is open source (MIT), written in TypeScript, and runs completely locally. If you jump between coding assistants and want to try it out, the repository is at: https://github.com/The-Untitled-... Would love to hear your thoughts, feedback on the architecture, or what other adapters you'd like to see supported next!

Wasla

Wasla – Universal agent sync

Wasla – Universal agent sync