A browser-based UI for building AI pipelines, chatting with models, processing images, and orchestrating workflows. Works with cloud APIs, local models, or both.
ModelPiper connects to AI providers to run inference. You can get started in minutes with any of these options:
macOS companion app that bundles local inference (llama.cpp, speech-to-text, text-to-speech, image upscaling), proxies cloud APIs securely, and manages model downloads.
Access 300+ models (GPT-4o, Claude, Gemini, Llama, Mistral, and more) through a single API key. Works directly in the browser — no install required.
Run open-source models entirely on your machine. Free, private, no API key needed.
ollama pull llama3.2)OLLAMA_ORIGINS=* ollama serveOpen My Connections from the sidebar to manage providers. Each provider becomes a configuration you can use across Pipelines, Chat, and other features.
OpenRouter works directly from the browser. Other cloud providers (OpenAI, Anthropic, Google Gemini) require the ToolPiper companion app to proxy requests securely.
Ollama and LM Studio run on your machine and connect over localhost. Models stay on your hardware — nothing leaves your network.
Any OpenAI-compatible API can be added as a custom provider. Set the host, port, path, and auth pattern to match your server.
Visual block-based editor. Chain LLM, audio, image, and RAG blocks together with automatic data flow between them. Blocks execute in sequence or parallel based on connections.
Multi-model chat interface. Supports streaming responses, system prompts, conversation history, and switching between any configured provider mid-conversation.
CoreML-powered image upscaling with side-by-side comparison. Requires ToolPiper for the upscaling backend.
Save entire pipeline configurations as reusable templates. Import and export your workspace including all settings and connections.
API keys are encrypted with AES-GCM and stored in your browser's localStorage. When ToolPiper is connected, keys are additionally stored in the macOS Keychain. Keys never leave your device.
Only the prompts and parameters you send to your chosen AI provider. ModelPiper has no backend server, no telemetry, and no analytics. Your data flows directly between your browser and the provider.
ModelPiper works standalone, but companion apps unlock additional capabilities:
Local AI gateway — inference engines, cloud proxy, model management, speech & audio
Screen capture, recording, GIF conversion, and live streaming
Multi-source audio mixer — mic, system audio, and per-app capture
Browser extension for image hover preview, AI upscaling, and page saving