ultrasync – semantic code search and memory for AI coding agents

Summary: ultrasync provides a semantic code search and memory layer for AI coding agents, enabling sub-millisecond lookups and persistent context across sessions. It builds an information retrieval index using pattern recognition and network-speed regex instead of relying solely on large language models.

What it does

ultrasync indexes codebases by recognizing patterns and maintains a lightweight IR without full file content indexing. It runs an MCP server with a transcript watcher to automatically extract context from chat transcripts and offers a CLI for index visibility.

Who it's for

It is designed for developers using AI coding agents like Claude and Codex who need efficient, persistent codebase understanding and memory across sessions.

Why it matters

ultrasync reduces reliance on costly LLM queries by enabling fast, local semantic search and memory, improving AI coding workflows without repo pollution or heavy indexing.