# Invincat

A terminal-native AI coding assistant built on DeepAgents CLI that inspects and edits project files, runs approved shell commands, keeps long-term memory across sessions, and can bridge scheduled or remote tasks in from WeCom messages.

Invincat runs from inside a project directory as a terminal UI, reading, editing and creating files, running shell commands under configurable approval, and using MCP tools and skills to extend what it can do. A plan mode lets you review and approve an execution plan before it makes any changes, and a separate memory agent captures what it learned after each task into hierarchical, long-term memory carried across sessions.

A model manager lets you register any provider, model name, API key and base URL interactively or through environment variables, with a separate optional model reserved for post-turn memory extraction. It can also create recurring or one-shot scheduled tasks described in natural language, and bridge remote WeCom messages into a project session.

It installs with pip as invincat-cli and requires Python 3.11 or newer.

Website: https://github.com/dog-qiuqiu/invincat