Supports optional meaning-based code search through a local Ollama embedding model, fused with keyword search results while keeping code on the local machine.
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., "@Quarry find where we compute the invoice total in my-app"
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.
Private code search for your machine, and for your AI agents. Index your repositories locally, search them by keyword, identifier or meaning, and give Claude Code (or any MCP client) a search_code tool that never sends your code anywhere.
Offline by default. Keyword search uses SQLite full-text search with BM25 ranking, built into Python, with nothing to install.
Understands code. It splits identifiers, so compute invoice total finds computeInvoiceTotal and max retry finds max_retry_count. Chunks follow functions and classes, and file paths and symbol names rank higher.
Optional meaning search with a local embedding model through Ollama (nomic-embed-text by default), fused with keyword results. Your code stays on your PC.
MCP server for Claude Code, Claude Desktop and other agents, with search_code, read_code, list_repos and reindex. It is read-only, confined to the folders you indexed, and refuses path tricks like ../.
Incremental. Re-indexing only re-reads changed files. It respects .gitignore, skips dependencies, build output and binaries, and never follows symlinks or junctions.
Get Quarry-0.1.1-windows.zip from Releases and unzip it anywhere. Quarry is a command-line tool: open a terminal in the Quarry folder and type Quarry.exe wherever this page says py -m quarry, for example Quarry.exe index D:\code\my-app. No Python needed.
Windows may say "Windows protected your PC" because the app is new: click More info, then Run anyway.
Prefer Python? With Python 3.9 or newer, download this repository and run py -m quarry in its folder.
Related MCP server: Codebase Copilot
py -m quarry index D:\code\my-app D:\code\other-repo
py -m quarry search "where do we retry failed uploads"
py -m quarry reposWith meaning search (needs Ollama and ollama pull nomic-embed-text):
py -m quarry index D:\code\my-app --meaning
py -m quarry search "how is the user's session refreshed" --meaningpy -m quarry mcp-config D:\code\my-appThis prints the exact claude mcp add ... line to run, and the JSON block for Claude Desktop or other MCP clients. After that, your agent can call search_code to find code in your private repositories without uploading anything.
Indexes live in %LOCALAPPDATA%\Quarry (~/.quarry elsewhere), one SQLite file per repository. Set QUARRY_HOME to move them. quarry forget <folder> deletes an index; it never touches your code.
py -m unittest discover -s testsThe core is MIT licensed. A future Pro edition (a desktop app with folder watching and a local "ask your codebase" chat) is planned as a paid add-on.
Made by jtclone22-coder: more free Windows tools are on that profile, and paid ones (DevSweep, Overnight, AssetAtlas) are on Gumroad.
This server cannot be deployed
Permission-aware onboarding MCP server: answers about a codebase, filtered by the caller's role.
Read-only MCP tools for AI agent discovery, structured resources, and NIULAI information.
The Cortex MCP server provides read-only access to real-time engineering context from the Cortex developer portal, allowing AI coding assistants to answer natural language questions about your organization's catalog (microservices, libraries, domains, teams, infrastructure), scorecards (engineering standards and best practices), initiatives (goals and deadlines), and Engineering Intelligence metrics. It includes tools for querying documentation, tracking personal entities, and accessing AI-assisted insights across the entire Cortex ecosystem.
Agent-native MCP server over the public saagarpatel.dev corpus. Read-only, stateless.
