FerroxLabs/wayland

What it solves

Wayland provides a unified command center for managing multiple AI CLI agents (such as Claude Code, Codex, Gemini, and Goose) from a single local application. It eliminates the need to juggle multiple separate command-line interfaces and prevents the "cold start" problem where agents forget context between sessions by providing shared, persistent memory across all tools.

How it works

Wayland runs locally on your machine, acting as a primary agent that can orchestrate other AI CLIs using the Agent Client Protocol (ACP). It uses a bundled engine called Fuigo and a native per-OS sandbox to execute shell commands and manage files securely. The system features a multi-tiered, SQLite-backed cognitive memory (divided into working, episodic, semantic, and procedural partitions) and can integrate with external tools via the Model Context Protocol (MCP).

Who it’s for

It is designed for developers and power users who use multiple AI coding and productivity agents and want a single, local-first interface to coordinate complex, multi-week projects without relying on cloud-based chat windows.

Highlights

• Unified Agent Control: Drive a dozen or more AI CLIs from one place using your own API keys.
• Self-Assembling Teams: Create teams of AI specialists led by a manager to work autonomously on a shared blackboard.
• Persistent Memory: Maintains state, files, and project context across sessions so you don't have to re-brief agents.
• Scheduled Workflows: Set recurring tasks in plain language to run on a schedule (cron).
• Local-First Security: Runs on your machine with native sandboxing (bubblewrap, sandbox-exec) to protect your files and shell.
• Cros-Platform Support: Available for macOS, Windows, and Linux, with an option for headless cloud self-hosting.