semble
by MinishLab
About
Semble is a fast, CPU-only code search library by MinishLab designed specifically for AI agents. It efficiently retrieves relevant code snippets via natural language queries, reducing token usage by ~98% compared to traditional grep-and-read methods. Requiring no API keys, GPUs, or external services, it indexes and queries codebases locally in milliseconds, achieving near-transformer retrieval accuracy. It functions seamlessly as an MCP server for instant integration with Claude Code, Cursor, and other agent workflows.
Features
- Sub-second CPU-only indexing and search
- Saves ~98% agent token consumption
- Zero setup without API keys or GPUs
- Native MCP server integration
- Near-transformer retrieval accuracy
Supported Platforms
desktop