semble

semble

A fast and accurate code search tool for AI agents, enabling efficient code retrieval.

MCPDevOpen source
Type
MCP
Transport
stdio
Open source
Yes
GitHub Stars
★ 5.4k
Source
mcp-github

Overview

Semble is a code search library designed specifically for agents, capable of instantly returning required code snippets, using approximately 98% fewer tokens than grep+read. It can index and search entire codebases in under a second and runs on CPU without requiring API keys, GPUs, or external services. It can be used as an MCP server, CLI tool, or dedicated sub-agent, suitable for any coding agent (e.g., Claude Code, Cursor, Codex). Through natural language queries, agents can immediately access relevant code snippets.

Capabilities

  • Fast indexing and searching of codebases
  • Uses less than 2% of the tokens compared to grep+read
  • Supports both local and remote repositories
  • No API keys, GPUs, or external services required
  • Supports multiple content types (code, documentation, configuration)

Use cases

Code reviewCode refactoringCode debuggingCode learning

Setup

Requires: Node 环境
After installing uv, run `uv tool install semble` and `semble install` to set up.

This information was compiled by AI from public sources and may contain inaccuracies — please refer to the source.

FAQ

How do I update Semble?

Run `uv tool upgrade semble` to upgrade, and `uv cache clean semble` to clear the cache.

What file types does Semble support?

By default, it supports code files; custom inclusion/exclusion rules can be defined via a .sembleignore file.

Related skills