HariAri - Run 12 agents side by side

Hariari provides a unified cockpit for developers who run multiple AI coding agents across several projects. It displays the status of each agent—running, waiting, or error—in a sidebar, and can issue visual and auditory alerts when an agent requires input. Users interact with agents through keyboard shortcuts or voice commands that are processed by a cloud speech‑recognition service, allowing hands‑free steering of tasks such as code generation, testing, or migration.

The environment is terminal‑native, embedding GPU‑accelerated xterm.js panes for each agent, and supports split‑pane layouts and layout presets to organize workspaces. It persists project configurations and can monitor agents in real time across Linux, macOS, and Windows platforms. Because it is not an editor or a terminal emulator, Hariari acts as a control plane that sits above existing CLI‑based AI agents, orchestrating their execution without altering their underlying behavior.

Hariari is released under the AGPL‑3.0 license and can be installed with a single npx command. It is experimental software aimed at “vibe coders” who need to coordinate several AI agents simultaneously, offering multi‑project visibility, configurable notifications, and optional voice control.