What file types can fs_read open?

It reads text and structured files including .txt, .md, .csv, .json, .xml, .log, .yaml and .toml, plus common code files like Swift, Python, JavaScript and Go. It's built for readable text, not binary formats.

Does fs_read upload my files to the cloud?

No. fs_read runs locally on your Mac through the LMCP server. File contents are never stored on any server or used for training. For desktop AI clients the data never leaves your machine; for web AIs only the content you request passes through an encrypted relay to your assistant.

Which AI assistants can use fs_read?

Claude Desktop, Cursor, VS Code (GitHub Copilot), Windsurf and Zed connect locally over stdio and configure automatically. ChatGPT, Claude.ai, Grok and Perplexity connect through the LMCP Cloud Relay connector.

Read Local Files with AI on Mac — fs_read

The fs_read tool lets your AI assistant open and read a text file directly from your Mac's filesystem. It handles plain text and structured formats — .txt, .md, .csv, .json, .xml, .log, .yaml, .toml — plus common source code files like Swift, Python, JavaScript, Go and more. Instead of copy-pasting file contents into a chat window, you just point the AI at the path and let it read.

A typical prompt looks like: "Read ~/Projects/api/config.yaml and tell me which environment variables are required." The AI calls fs_read, pulls the real file off your disk, and answers from the actual content — no manual copy, no stale paste.

Which AI agents work?

LMCP exposes fs_read to any MCP-capable assistant. Desktop clients auto-configure over local stdio; web AIs connect through the LMCP Cloud Relay connector.

Desktop (local stdio, auto-configured): Claude Desktop, Cursor, VS Code (GitHub Copilot), Windsurf and Zed.
Web (LMCP Cloud Relay connector): ChatGPT, Claude.ai, Grok and Perplexity.

Once LMCP is installed and your client is restarted, fs_read appears automatically alongside the other 150+ tools.

Automation

Reading a file is rarely the end goal — it's the first step in a chain. fs_read pairs naturally with other LMCP tools: read a CSV of contacts and draft personalized emails, read a Markdown changelog and post a summary to Slack or Teams, read a JSON export and create matching calendar events or reminders, or read a YAML config and explain a deployment to a teammate over Messages. Because every tool runs on the same local server, the AI can move from "read this file" to "now act on it" in a single conversation.

Context

The biggest win is that the AI works with your real Mac data, not a generic guess. When you ask it to debug a .log file, refactor a code snippet, or validate a config.toml, it reads the exact bytes on your disk — current, accurate, and specific to your machine. No upload, no syncing, no "please paste the file again." The assistant sees what you see.

Productivity

Copy-pasting long files into a chat is slow and error-prone — you truncate, you miss lines, you lose formatting. With fs_read the AI grabs the whole file in one call. Reviewing a 400-line config, summarizing a CSV report, or explaining an unfamiliar source file drops from minutes of manual shuffling to a single sentence. Multiply that across a workday of "what's in this file again?" moments and the time saved adds up fast.

Privacy & GDPR

Everything happens on your Mac. fs_read opens files locally through the LMCP server running on your own machine — file contents are never uploaded to LMCP servers, never stored in the cloud, and never used for training. For desktop clients the data path never leaves your computer; for web AIs only the specific content you request flows through the encrypted relay to your chosen assistant. This is privacy by architecture — see GDPR-compliant by architecture for details.

Ready to give your AI read access to your files? Download LMCP — it's free and native to macOS.