wtf is lookingglass?

Looking Glass is a command line tool for coding with a local AI assistant, designed around sessions that can persist and continue working even after you close the terminal. Instead of treating each AI conversation as a one-off chat, Looking Glass saves the full conversation, tool history, and workspace state to a local SQLite database, so a session can be reopened later and continue exactly where it left off. The standout feature is its built-in scheduler. You can schedule a session to wake up later and take another AI turn automatically, for example to check whether a deployment finished, follow up on test results and fix failures, run a recurring code review, or continue debugging after new logs appear. These scheduled turns are not separate stateless jobs, they can inspect files that changed, run commands, edit code, and write results back into the same ongoing session. The scheduler runs as a background service independent of the terminal, using systemd on Linux and Task Scheduler on Windows. Looking Glass works with an interactive terminal chat interface as well as one-shot command line prompts, and it can run multiple worker agents concurrently, each with its own model and reasoning settings, so a stronger model can coordinate work while cheaper or faster models handle parallel tasks. It connects to any OpenAI-compatible model gateway, including LM Studio running locally or other compatible servers, rather than requiring a specific cloud AI provider. The tool is built for a single local operator rather than as a hosted or multi-user service. It requires Linux or native Windows, Node.js version 22.19.0 or newer, and the ripgrep search tool available on your system path, macOS is not currently supported. It can be installed globally through npm, or built from source by cloning the repository and running the project's build scripts. Configuration, including which model and gateway to use, is controlled through a JSON or JSONC settings file that can be set globally or per project.