loaditout.ai
SkillsPacksTrendingLeaderboardAPI DocsBlogSubmitRequestsCompareAgentsXPrivacyDisclaimer
{}loaditout.ai
Skills & MCPPacksBlog

gitdex-semantic-search

MCP Tool

BenDavies1218/gitdex-semantic-search

Local code indexer with MCP-based semantic search. Parses Git repos with tree-sitter, embeds via Ollama, stores in Qdrant, and exposes search as MCP tools for Claude Code, Cursor, and other LLM dev tools.

Install

$ npx loaditout add BenDavies1218/gitdex-semantic-search

Platform-specific configuration:

.claude/settings.json
{
  "mcpServers": {
    "gitdex-semantic-search": {
      "command": "npx",
      "args": [
        "-y",
        "gitdex-semantic-search"
      ]
    }
  }
}

Add the config above to .claude/settings.json under the mcpServers key.

About

gitdex

[](https://github.com/BenDavies1218/gitdex-semantic-search/actions/workflows/ci.yml)

Local semantic code search for LLM-assisted development.

Gitdex indexes your Git repositories using tree-sitter AST parsing and vector embeddings, then exposes semantic search via Model Context Protocol (MCP) tools. Point Claude Code, Cursor, or any MCP-compatible tool at your codebase and search it with natural language.

gitdex index ./my-project        # Index the repo
claude mcp add gitdex -- \       # Register with Claude Code
  gitdex serve ./my-project
How It Works
  Repository          gitdex              Services
  ──────────    ──────────────────    ──────────────────

  .git/repo  ──>  Walk files          Ollama
                  (gitignore-aware)   (nomic-embed-text)
                       │                    ^
                       v                    │
                  Chunk with           Embed chunks ──>  Qdrant
                  tree-sitter                           (vector DB)
                       │                                    ^
                       v                                    │
                  MCP Server  ──────── search ──────────────┘
                  (stdio)
  1. Walk — discovers files respecting .gitignore, skips binaries and lock files
  2. Chunk — parses supported languages into functions, classes, and methods via tree-sitter; falls back to line-based chunking for other text files
  3. Embed — generates 768-dimensional vectors via Ollama (nomic-embed-text) with bounded concurrency
  4. Store — upserts vectors to Qdrant with payload indexes for fast filtered search
  5. Search — MCP server embeds your query and performs cosine similarity search over the index

Subsequent runs use incremental indexing — only files changed since the last indexed com

Tags

claude-codecode-searchdeveloper-toolsmcpollamaqdrantrustsemantic-searchtree-sittervector-search

Reviews

Loading reviews...

Quality Signals

0
Installs
Last updated19 days ago
Security: AREADME

Safety

Risk Levelmedium
Data Access
read
Network Accessnone

Details

Sourcegithub-crawl
Last commit3/27/2026
View on GitHub→

Embed Badge

[![Loaditout](https://loaditout.ai/api/badge/BenDavies1218/gitdex-semantic-search)](https://loaditout.ai/skills/BenDavies1218/gitdex-semantic-search)