Agent Workbench is a local-first web workspace for managing coding-agent work across multiple projects and sessions. It starts with Gemini CLI support and adds a dashboard for session status, isolated worktrees, code review, snapshots, apply/sync flows, terminal attach, voice input, screenshot paste, and draft PR delivery. It is built for developers who want to run several agent tasks in parallel without losing control of their repos.
A simple, lightweight shell script to interact with OpenAI or Ollama or Mistral AI or LocalAI or ZhipuAI from the terminal, and enhancing intelligent system management without any dependencies(pure shell).