wiggins-j/errorta_app

Errorta is a desktop application, built with Tauri, React, and a Python backend, that lets you assemble a team of AI coding agents to actually write and ship software for you. Instead of talking to a single AI assistant, you set up a project manager agent alongside programmer and reviewer agents, each of which can run on a different AI model or subscription of your choosing, including Claude, OpenAI, Google models, or a locally run Ollama model. Given a goal or an existing repository, the project manager agent breaks the work into a task queue and assigns it across the team. Programmer agents write code on separate git branches, tests run in a sandboxed environment, and reviewer agents check the resulting code changes before anything is merged. When work is imported from an existing project, the tool can read the README and code to suggest a goal and definition of done for you to approve, and any finished work comes back as a GitHub pull request rather than being applied directly. Beyond the coding team feature, Errorta also supports multi-model group discussions where several AI models weigh in together, answering questions using your own documents while noting how confident each answer is, and can be controlled remotely from an iPhone on the same network. The tool can run entirely locally with no internet connection using Ollama models and a local document store, or it can use cloud AI providers and existing subscriptions, mixed and matched per project or per team member. It ships as a full desktop app you build from source with Node.js, Rust, and Python, or as a separate command line tool called errorta that can drive the same coding team from a terminal, useful for scripting or continuous integration. A prebuilt CLI binary is available for Apple Silicon Macs through Homebrew.