Connects AI assistants to Hawkeye's local search engine for navigating large codebases without uploading code. Exposes nine MCP tools that let Claude or Copilot find references to classes, functions, symbols, assets, and localization keys across indexed projects. Runs entirely on-premises with no AI-based indexing. The main win is token efficiency: instead of reading files blindly, your assistant queries Hawkeye first and gets back precise locations, cutting context usage by roughly 80%. Requires both Hawkeye and Hawkeye AI Bridge installed locally. Works with Claude Desktop (via .mcpb), Claude Code, VS Code Copilot in Agent mode, Cursor, and OpenCode. Includes a skill file so Claude knows when to reach for search instead of guessing.