kapala-solutions/ravenspire

Ravenspire turns your AI coding agents into characters in a pixel-art role-playing game. If you use tools like Claude Code or OpenAI Codex, each work session appears as a hero in a fantasy guild world. The task the agent is working on becomes a quest, and when the agent is actively doing something, it battles a monster whose size reflects the difficulty of the job. The biggest active project spawns a world boss with its own health bar. Behind this game layer, Ravenspire functions as a real operations dashboard that tracks costs, response times, and work history. The project exists because running many AI agents at once creates two problems. First, it is hard to know which agent is working, which is stuck, and which is waiting for your input, and what all of it costs. Second, standard dashboards are dull, so people stop checking them. Ravenspire addresses both issues by wrapping genuine monitoring data in a 16-bit style game that makes you want to pay attention. The main view is the quest world, where heroes have classes, levels, and experience points derived from token usage. You can click any hero to see their current quest, their actual last message, career statistics, and real tool usage displayed as skill bars. A control panel tab groups sessions by workspace and shows model, token breakdown, API cost, and labor value for each agent. A history tab tracks lifetime totals and trends. Waiting agents display the actual question they are asking you, with a timer showing how long they have been waiting. Setup requires Node 18 and at least one supported coding agent. You clone the repository, run install, run a setup script that wires hooks into your agent configuration files, and start the server. The setup script backs up your existing configuration and can be re-run safely. On macOS and Linux, you also need PowerShell 7 installed. Optional features include phone push notifications through a free service called ntfy, a desktop app install mode, and automatic startup on Windows login. The whole thing runs on plain Node with no build step and a single dependency. All state is stored in human-readable files. The full README is longer than what was shown.