Where does fs_search start looking?

By default it searches from your home directory. You can point it at any root folder — like Documents, Downloads or a specific project directory — and it searches recursively from there.

Does the search match partial names?

Yes. Matching is case-insensitive and partial, so searching for 'budget' finds 'Q3-Budget-final.xlsx', 'budgeting.md' and 'BUDGET' alike. You don't need the exact filename.

Do my filenames get sent to the cloud?

No. fs_search runs entirely on your Mac. The folder walk happens locally and only the matching paths your AI asked for are returned to the conversation — nothing is uploaded or stored on a server.

Find Files by Name on Mac with AI

fs_search lets your AI assistant find files and folders on your Mac by name. It does a case-insensitive, partial-name match starting from a root directory you choose, and defaults to your home folder if you don't specify one. Instead of clicking through Finder or remembering exactly where you saved something, you just describe it. For example, you can type: "Find every file on my Mac with 'invoice' in the name from this year." The AI calls fs_search, walks your folders locally, and hands back the matching paths in seconds.

This is part of LMCP, a free, native macOS MCP server that gives AI assistants access to your real Mac apps and files locally — with 150+ tools across Mail, Calendar, Finder, Notes and more. There are no API keys and nothing is copied to a server. Download LMCP to get started.

Which AI agents work?

You can use fs_search from almost any modern AI client:

Desktop apps — Claude Desktop, Cursor, VS Code (GitHub Copilot), Windsurf and Zed auto-configure through a local stdio connection. Install LMCP, restart the client, and the tools appear.
Web AIs — ChatGPT, Claude.ai (web), Grok and Perplexity connect through the LMCP Cloud Relay connector, which bridges the web assistant to the server running on your Mac.

Either way the search itself always executes on your machine — the AI only ever sees the file paths it asked for.

Automation

The real power shows up when fs_search chains with other LMCP tools. Find a file, then act on it in one conversation:

"Find the latest version of my pitch deck, then attach it to an email to my co-founder." — fs_search → Mail.
"Locate all the PDFs in my Downloads with 'contract' in the name and open the most recent one." — fs_search → Finder.
"Search my Projects folder for files named 'budget' and summarize what's inside the spreadsheet." — fs_search → file read.

You describe the outcome; the AI sequences the tools.

Context

Because fs_search reads your actual file system, the AI works with your real data — your true folder structure, the files you saved last week, the naming conventions you actually use. There's no need to upload anything or paste paths. Ask "where did I save the onboarding doc?" and the assistant answers from your live Mac instead of guessing.

Productivity

Hunting for a file you saved months ago is one of those small, constant time-sinks — opening Finder, second-guessing the folder, trying three spellings. With fs_search you ask once and the AI returns the path instantly, even on a partial or fuzzy name. Multiply that across a workday and it adds up to real saved minutes, and it keeps you in the flow of whatever you were actually doing.

Privacy & GDPR

Everything fs_search does runs locally on your Mac. The directory walk happens on your machine, no file index is uploaded, and nothing is stored on any LMCP server. Your filenames and folder structure never leave your computer. This local-first design makes LMCP GDPR-compliant by architecture — there's no cloud copy of your data to leak, subpoena or breach.