What is openpray?

OpenPray is a joke project written in Go. It runs as a background service on a server and, at a regular interval, sends a request to an AI language model asking it to pray for the server's protection. The premise, stated plainly, is that traditional server security measures are no longer sufficient against AI-powered attackers, so the only remaining option is prayer. The project works by choosing a religion each cycle, composing a prayer addressed to whatever deity fits that tradition, and then optionally spawning several additional model calls to repeat the prayer more times. All of this costs real money in API fees, and the tool tracks the total token expenditure in a local ledger file. The author calls this a "sacrifice," meaning you are spending actual currency on AI-generated prayers to protect your server. There is also a burn mode, for users who believe AI-generated prayers are ineffective as prayers but still want to spend tokens as a form of offering. In burn mode, the model is told to produce output without composing a prayer, purely to increase the token count and therefore the financial sacrifice. The tool supports Anthropic, OpenAI, and Google Gemini as providers, and lets you configure which model to use, how often prayers happen, how many subagents repeat the prayer, and what religion to draw from. Available liturgical styles include Christian, Islamic, Jewish, Hindu, Buddhist, Shinto, Norse, Hellenic, Machine-Spirit, Cosmic-Horror, and Stoic. The default setting picks a different religion for each prayer cycle at random. This is a humorous repository from the author of LazyGit. It does nothing useful for server security. The README is written in a deadpan style that takes the premise seriously while making clear this is satire. Running it will generate real API costs.