FlowDown5.1.8

FlowDown lets users switch between AI services, connect MCP tools, and run local models from a native agent app. It supports vision-capable models, audio messages sent as attachments to compatible models, file and document attachments, real-time web search, and reusable prompt templates. Its tool-calling features can search the web and process documents, with Gemini Flash recommended for the strongest tool-calling experience. The app works with models using the OpenAI API format, including OpenAI, Claude through OpenRouter, Alibaba Cloud, ByteDance, Ollama, and MLX local deployments, as well as custom compatible services.

Deep integration with system Shortcuts supports workflow automation, while model configurations can be imported and exported. Users can also supply their own API keys for OpenAI-compatible services, making FlowDown useful for configuring several providers or combining hosted and on-device models. As productivity software, it suits research, drafting, document analysis, web-informed questions, and recurring prompt-based tasks.

Conversation data is stored locally, FlowDown does not collect or retain conversation content, and requests go directly to AI providers without intermediaries. Its native app lineup includes a full-featured desktop edition, an iPhone app for iOS 17 or later, and a tablet edition adapted for iPadOS. The code is open source, providing transparency into the application.

The desktop release requires macOS 14 Sonoma or later and supports both Intel and Apple Silicon. Version 5.1.8 is freeware and includes core chat, model configuration, and local storage. Detailed configuration guides and community support are available for setup and provider-specific use.