Semble

Name: Semble
Availability: InStock
Author: minishlab

Supports UI

by minishlab

Fast, accurate code search for AI agents using semantic and lexical retrieval with 98% fewer tokens than grep.

0 stars

2 views

Works in:CursorCodex

Exposes:Tools

View on GitHub Docs

What it does

Semble is a high-performance code search library designed specifically for AI agents. It allows agents to instantly find relevant code snippets in any repository (local or remote) using a hybrid of semantic embeddings and BM25 lexical search, significantly reducing token consumption and latency compared to traditional grep-and-read workflows.

Tools

search: Find code using natural-language or code queries. Supports git URLs or local paths.
find_related: Returns semantically similar code chunks based on a specific file path and line number.

Installation

Install using uv:

uvx --from "semble[mcp]" semble

For Claude Desktop, add to claude_desktop_config.json:

{
  "mcpServers": {
    "semble": {
      "command": "uvx",
      "args": ["--from", "semble[mcp]", "semble"]
    }
  }
}

Supported hosts

Confirmed support for Claude Code, Cursor, Codex, and OpenCode.

Quick install

uvx --from "semble[mcp]" semble

Information

Pricing: free
Published: 5/1/2026
stars

Related Apps

Choose your AI client and follow the steps below.

Codex

Add to ~/.codex/config.toml: [mcp_servers.semble] command = "uvx" args = ["--from", "semble[mcp]", "semble"]

Cursor

Add to ~/.cursor/mcp.json: { "mcpServers": { "semble": { "command": "uvx", "args": ["--from", "semble[mcp]", "semble"] } } }

Claude Desktop

{
  "mcpServers": {
    "semble": {
      "command": "uvx",
      "args": ["--from", "semble[mcp]", "semble"]
    }
  }
}