A local structural code index shared across AI coding sessions

Most coding agents start every session by rediscovering the repository using grep and file reads. We built Sem to calculate that structure locally once and reuse it: - Functions, classes, methods and types. - Callers and dependencies. - Entity-level history and diffs. - Impact analysis. - Bounded context for agents. - No source-code upload required for local use. - CLI, JSON and MCP access. - One native binary. The index persists across sessions, so it does not require a permanent background daemon. On Sem’s own repository, an initial query took approximately 185 ms and a warm query took around 7 ms. Install and source: https://github.com/Ataraxy-Labs/sem The longer-term goal is to prebuild these indexes into self-hosted development containers, allowing every local agent to start with a warm understanding of the repository. Would people here prefer this as a standalone binary, a Docker image, a Dev Container Feature, or an integrated service shared across multiple machines?