Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Local Search find where API keys are validated in my indexed code"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A macOS app and local MCP service for searching your code, documents and images with natural language, using the EmbeddingGemma 2 model. Indexing and search run on your Mac; one background service shares the model across folders and assistants.
Requires Apple Silicon and macOS 14 or newer. To let Claude Code or Codex do it for you, paste:
Install Local Search by following the Install section of https://github.com/btroussel/SemanticSearchMCP#install, then connect it to yourself as an MCP server. Ask me whenever a step needs me in macOS or in the app.Download Local-Search-<version>.dmg from the latest release and drag Local Search to Applications.
Open it. The app is not notarized, so macOS blocks the first launch: choose Open Anyway in System Settings → Privacy & Security.
Click Install. The app downloads Python with its libraries (about 1.7 GB) and the model (1.5 GB) once, or uses a model folder you already have.
Add the folders to index in the app. Nothing is indexed by default.
Connect an assistant, then start a new session:
# Claude Code
claude mcp add --scope user --transport stdio local-search -- \
"$HOME/Library/Application Support/Local Search/runtime/venv/bin/code-search" mcp --general \
--url http://127.0.0.1:8766 \
--token-file "$HOME/Library/Application Support/Local Search/access.key"
# Codex
codex mcp add local-search -- \
"$HOME/Library/Application Support/Local Search/runtime/venv/bin/code-search" mcp --general \
--url http://127.0.0.1:8766 \
--token-file "$HOME/Library/Application Support/Local Search/access.key"The app's Connect an assistant panel shows the same commands. By default, an assistant can only search its own project folder; see MCP.
Everything the app installs or stores lives in ~/Library/Application Support/Local Search/. To uninstall, delete the app and that folder.
The network is used only during setup. Snippets and image previews that a cloud assistant requests are sent to its provider like any other tool output; see privacy.
Related MCP server: Apple Tools MCP
Requires uv and the Xcode command-line tools.
uv sync --extra dev
.venv/bin/python scripts/build-mac-app.py # standalone app, same first-launch setup
open '.code-search/Local Search.app'See development for the --dev build linked to this checkout, the CLI and tests.
Topic | Guide |
Using the app, settings, supported files and limits | |
Connecting assistants, project access, tools and API | |
What is accessed, stored and sent | |
How it works, code map, builds, CLI and checks |
Coding agents working on this repository should read AGENTS.md.
Apache-2.0. The EmbeddingGemma 2 checkpoint is downloaded separately and is also distributed under Apache-2.0.
This server cannot be deployed
Persistent memory and knowledge graph for AI assistants — keyword + vector + graph search.
Search your knowledge bases from any AI assistant using hybrid RAG.
Project memory, semantic code search, and grounded agent context.
Token-efficient search for coding agents over public and private documentation.
