Semble
A code search library built for AI agents that returns exact relevant code snippets using roughly 99% fewer tokens than a typical grep-and-read search.
Semble answers a natural-language query, such as how authentication is handled, by returning only the relevant code snippets rather than requiring an agent to grep for matches and then read whole files. Indexing and searching a full codebase end-to-end takes under a second for most repositories, and the project reports matching the retrieval quality of a code-specialized transformer model while indexing about 380 times faster and querying about 17 times faster, with everything running on CPU and no API keys, GPU, or external services required.
An interactive installer detects installed coding agents such as Claude Code, Codex, and OpenCode and lets a user choose which integrations to enable: an MCP server so an agent can call Semble as a tool directly, instructions added to AGENTS.md or CLAUDE.md, or a dedicated semble-search sub-agent. An unattended install mode supports sandboxed or scripted environments.
It is written in Python and distributed under the MIT license.