--- title: "Local MCP Server on Mac: 300+ AI Tools in One Install" description: "ToolPiper is a 300+ tool MCP server for macOS. Local inference, browser automation, voice, vision, desktop control, and testing from one native app." date: 2026-03-27 updated: 2026-05-14 author: "Ben Racicot" tags: ["MCP", "Model Context Protocol", "Developer Tools", "Privacy", "macOS", "Claude Code", "AI Tools", "Apple Silicon"] type: "article" canonical: "https://modelpiper.com/blog/mcp-server-local-mac/" --- # Local MCP Server on Mac: 300+ AI Tools in One Install > ToolPiper is a 300+ tool MCP server for macOS. Local inference, browser automation, voice, vision, desktop control, and testing from one native app. ## TL;DR Claude Code, Cursor, and Windsurf are only as powerful as the tools they can reach. ToolPiper is a native macOS MCP server that gives your AI assistant over 300 tools spanning local LLM inference, TTS, STT, browser automation, desktop control, web scraping, testing, and video production. One install, one command to connect, everything runs on your hardware. Claude Code, Cursor, and Windsurf are only as powerful as the tools they can reach. MCP lets you extend them, but most MCP servers do one thing. A file reader. A database connector. A web searcher. ToolPiper is a native macOS app that exposes over 300 MCP tools covering local inference, browser automation, voice, vision, web scraping, desktop control, testing, and video production. One install, everything on your hardware. This article covers what MCP is, why local matters, and what those over 300 tools actually do. ## What is MCP and why should you care? MCP stands for Model Context Protocol. It's an open protocol from Anthropic for connecting AI assistants to external tools. A server exposes tools (functions the AI can call), a client like Claude Code discovers those tools and invokes them, and JSON-RPC handles the wire format. Think of it as a USB port for AI. Before MCP, every assistant had its own plugin format and integration story. MCP standardizes that. Write a tool once, any MCP client can call it. There are already MCP servers for Playwright, filesystem access, databases, GitHub, Slack, and dozens more. Most of them share a common limitation. ## Why does it matter that an MCP server runs locally? Most MCP servers are cloud wrappers. They proxy your requests to external APIs, which means your data still leaves your machine. The MCP server for OpenAI sends your prompts to OpenAI. The protocol is local, but the execution isn't. A local MCP server is different. The tools run on your hardware. When you ask your AI assistant to transcribe audio, summarize a document, or chat with a local LLM, the computation happens on your Mac's Neural Engine and Metal GPU. Nothing crosses a network boundary. **Privacy**: your code, documents, and voice recordings never leave your machine. **Speed**: no round trip to a data center, no cold starts. **Availability**: works on a plane, on a train, and when your ISP goes down. ## What is ToolPiper? ToolPiper is a native macOS application that bundles six inference backends (llama.cpp, Apple Intelligence, FluidAudio STT/TTS, MLX Audio TTS, Apple Vision OCR, CoreML) behind a single HTTP gateway on localhost. It's the local engine behind [ModelPiper](https://modelpiper.com), the visual AI pipeline builder. It's also a full MCP server. One install gives your AI assistant over 300 tools spanning local inference, browser automation, desktop control, testing, web scraping, video creation, and more. Setup takes 30 seconds. ## How do you connect ToolPiper to Claude Code? Install ToolPiper from [modelpiper.com](https://modelpiper.com/docs/toolpiper). Then run one command: ``` claude mcp add toolpiper -- ~/.toolpiper/mcp ``` Restart Claude Code. That's it. No npm install, no Docker, no Python environment. ToolPiper places a symlink at `~/.toolpiper/mcp` pointing to the native binary inside the app. It updates automatically when you update ToolPiper. Two MCP transports are supported. **stdio** is the universal one - Claude Code, Cursor, and most CLI tools use it. A lightweight Swift executable reads JSON-RPC from stdin and bridges to ToolPiper's local HTTP API. **Streamable HTTP** is the newer transport for web-based clients, served at `POST localhost:9998/mcp`. Both transports expose identical tool definitions. For Cursor, Claude Desktop, Windsurf, VS Code (with Cline or Continue), Zed, or Aider, each editor has a slightly different registration syntax. See [MCP Setup on Mac](/blog/mcp-setup-guide-mac) for the full per-editor walkthrough, or [Connect ToolPiper on Mac](/blog/connect-toolpiper) for the broader integration hub covering OAuth and the local HTTP API alongside MCP. ## What are the 300+ tools? The tools break down into eleven categories. Some have a handful of focused tools. Others, like desktop control, have dozens. **Local AI Inference (12 tools).** The foundation. `chat` runs prompts through a local LLM. `transcribe` converts audio to text on-device. `speak` synthesizes speech. `ocr` extracts text from images and PDFs using Apple Vision. `embed` generates vector embeddings for RAG pipelines. `analyze_image` and `analyze_text` handle multimodal and text analysis. The remaining tools manage the model lifecycle: listing what's available, checking download status, loading and unloading from memory. All inference runs on your Mac's Neural Engine and Metal GPU. **Knowledge Base (3 tools).** Local retrieval-augmented generation. Index a document collection with `rag_ingest`, search it with `rag_query` using hybrid HNSW vector + BM25 keyword retrieval with semantic chunking, and list your indexed collections with `rag_collections`. Everything stays on-device. **Browser Automation (15 tools).** Full Chrome DevTools Protocol control using the accessibility tree instead of CSS selectors. `browser_snapshot` captures page state. `browser_action` clicks, types, and fills forms, returning an AX diff of what changed. Seven assertion types with polling via `browser_assert`. Console reading, interaction recording, network monitoring, cookie and storage management, Web Vitals measurement, JS/CSS coverage tracking, JavaScript execution, request mocking, passkey simulation, and form autofill round out the set. **Web Scraping (2 tools).** `scrape` extracts content from web pages in seven formats (markdown, plain text, readability, AX tree, HTML, links, screenshot) using a real browser with readiness detection for 16 JavaScript frameworks. `browser_detect` identifies which frameworks a page uses. **Testing (10 tools).** PiperTest is a visual, AX-native test format with self-healing selectors. Six tools handle test session CRUD, execution, and export to Playwright or Cypress code. Four additional Sieve tools analyze and auto-repair broken selectors across your test suite. **Pose and Motion Capture (3 tools).** Real-time skeleton tracking using Apple Vision. Single-image pose estimation, WebSocket streaming at 60fps, and four output formats including a zero-allocation compact binary format at 236 bytes per frame. **Desktop Control (29 tools).** Full macOS system control through ToolPiper. Window management with snap layouts. Keyboard and mouse simulation. Volume and display brightness. Wi-Fi, Bluetooth, Dock, desktop, Spaces, Focus modes, media playback. Power and process management. Finder operations. App interaction (list running apps, launch, capture snapshots, run assertions). Accessibility and appearance settings. Notifications, Calendar, Contacts, Reminders, Location, Shortcuts, system defaults, and storage info. **Video Production (18 tools).** AI-driven video creation from screenplay to final render. Project management, media import, screenplay and composition editing, timeline editing, dry-run rehearsal, screen recording, rendering, AI narration, preview, timeline export, clip management, recording settings, and PiperSR video upscaling. **Social and Research (10 tools).** GitHub repository analysis and cross-repo comparison. Hacker News trending stories and search. Reddit search and post retrieval. X/Twitter browsing and composition. YouTube transcript extraction. **Files, Git, and Utilities.** The remaining tools cover file operations (read, write, create, delete, list, directory picker, shell commands), Git integration (status, diff, log, commit, push, checkout), image upscaling with benchmarking, voice cloning, live streaming control, content queue management, OAuth connection management, and API discovery. ## How does this compare to other MCP servers? Most MCP servers are single-purpose. Here's how ToolPiper compares to the alternatives you'd need to combine for equivalent coverage. ToolPiper Playwright MCP Filesystem MCP Browser MCP Number of tools 303 12 5 8 Local AI inference Yes (LLM, TTS, STT, OCR, embeddings) No No No Browser automation Yes (CDP + AX tree) Yes (Playwright) No Yes (basic) Desktop control Yes (29 system actions) No No No Testing Yes (PiperTest + export) Partial No No Web scraping Yes (7 formats, framework-aware) No No No Voice/Audio Yes (transcribe, speak, clone) No No No Video production Yes (18 tools, screenplay to render) No No No Setup One app install npm install npm install npm install The difference isn't just tool count. ToolPiper's tools compose. A single workflow can transcribe audio, pass the text to a local LLM, query your RAG index for context, and automate a browser action based on the result. One server process, shared model state, shared authentication. ## What are the honest limitations? **macOS only.** ToolPiper requires Apple Silicon (M1 or later). The inference backends depend on Metal GPU, Neural Engine, and Core Audio frameworks that only exist on macOS. No Windows, no Linux. **Some tools require ToolPiper Pro.** All inference tools and all read-only tools are free. Test mutations (save, delete), developer tokens, and some advanced features require ToolPiper Pro at $9.99/month. **Browser tools need Chrome.** Browser automation uses Chrome DevTools Protocol. You need Chrome or Chrome Dev installed. Safari, Firefox, and Arc aren't supported for automation. **Over 300 tools is a lot of tools.** Tool selection quality depends on the AI client. Claude Code handles large tool sets well because Anthropic designed MCP with this in mind. Other clients may struggle when presented with over 300 options. ToolPiper's tool descriptions are written to help AI models choose correctly, but results vary across clients. ## Does this work with Cursor and Windsurf? Yes. Any MCP-compatible client can connect. For stdio-based clients like Claude Code and Cursor, use the symlink at `~/.toolpiper/mcp`. For HTTP-based clients, point to `http://localhost:9998/mcp`. Tool definitions are identical across both transports. For Claude Code: ``` claude mcp add toolpiper -- ~/.toolpiper/mcp ``` For Cursor and other JSON-config clients, add ToolPiper to your MCP server list pointing to the same binary. ## Do all 300+ tools run locally? All inference runs locally. Chat, transcribe, speak, embed, OCR, image analysis, pose detection, and video upscale execute on your Mac. Your prompts, audio, and documents never leave the machine. Social and research tools (GitHub, Hacker News, Reddit, X, YouTube) make network requests to fetch public data. Browser automation interacts with whatever page is loaded in Chrome, which may involve network traffic. Desktop control operates entirely on your local system. ## Is this free? ToolPiper's free tier includes all inference tools, all browser automation tools, all read-only operations, and MCP access through both transports. You can use it as a full MCP server without paying anything. ToolPiper Pro is $9.99/month. It adds test mutations, developer tokens, advanced model management, and priority features. The MCP server itself isn't gated behind Pro. ## How do you see what tools are available? From Claude Code, ask your assistant to call `status`. It returns server health, loaded models, and available capabilities. Call `models` to see which AI models are downloaded and ready. ToolPiper serves a full OpenAPI spec at `http://localhost:9998/v1/openapi.json` documenting every REST endpoint. The MCP tool definitions include detailed descriptions and JSON Schema for every parameter. For a deeper look at how the MCP server is built, including the two-transport architecture and shared-definition pattern, see [Building Over 300 MCP Tools in Swift](/blog/building-over-140-mcp-tools-in-swiftswift). _This article is part of the [local-first AI on macOS](/blog/local-first-ai-macos) series. For the visual testing tools, see [Visual Testing, No Code](/blog/visual-testing-no-code-mac)._ ## FAQ ### What hardware do I need to run ToolPiper? ToolPiper requires a Mac with Apple Silicon (M1 or later) running macOS 26 or newer. 8GB of RAM is the minimum, but 16GB or more is recommended if you plan to run larger language models (7B+ parameters). All inference runs on the Neural Engine and Metal GPU built into your Mac's chip. No external GPU or additional hardware required. ### Does ToolPiper work with Cursor and Windsurf? Yes. ToolPiper supports two MCP transports: stdio for CLI tools like Claude Code and Cursor, and Streamable HTTP for web-based clients. Any MCP-compatible client can connect. For stdio clients, point to `~/.toolpiper/mcp`. For HTTP clients, use `http://localhost:9998/mcp`. The tool definitions and behavior are identical across both transports. ### Do all 300+ tools run locally? All inference tools (chat, transcribe, speak, embed, OCR, image analysis, pose detection, video upscale) run entirely on your Mac's hardware. Your prompts, audio, and documents never leave the machine. Social and research tools (GitHub, Hacker News, Reddit, X, YouTube) make network requests to fetch public data. Browser automation interacts with whatever page is loaded in Chrome. Desktop control operates on your local macOS system. ### Is ToolPiper free? The free tier includes all inference tools, all browser automation tools, all read-only operations, and full MCP access through both transports. ToolPiper Pro ($9.99/month) adds test mutations, developer tokens, advanced model management, and priority features. The MCP server itself is not gated behind Pro. ### Can I use my own models with ToolPiper? Yes. ToolPiper runs any GGUF-format model through its llama.cpp backend. You can download models directly from HuggingFace through the app or use the `download_model` MCP tool. Custom models appear alongside ToolPiper's curated model list and are available through all the same inference tools.