Invincat is a Python-based terminal AI coding assistant designed to work directly inside local repositories. It can read and write files, execute commands, perform web searches, and coordinate multiple agents to participate in real development workflows rather than just answer questions in chat. Its biggest highlight is memory management. Instead of relying on users to manually maintain context, Invincat uses an independent Memory Agent to review each completed task, extract valuable information, summarize user preferences, project conventions, coding habits, and historical decisions, then organize them into structured long-term memory. The system separates user-level and project-level memory to avoid mixing personal preferences with project facts. It also supports scoring, hot/cold memory tiering, archiving, and retiering, so frequently used or important memories are prioritized while less relevant ones are moved into colder storage. It can also deduplicate memories, detect conflicts, and delete incorrect, outdated, or contradictory entries. The result is an auditable, maintainable, self-cleaning memory system that gives the agent stronger continuity across sessions.