A native workspace for many AI models
BoltAI brings AI providers, local models, conversations, and reusable tools into a native Mac application built with SwiftUI and AppKit. The official site says users can switch among more than 300 models and connect services including OpenAI, Anthropic, Google, Mistral, Azure, Bedrock, OpenRouter, Ollama, and LM Studio. The goal is to keep model choice separate from the structure of the work.
Workflows beyond a single chat
The app supports projects, multi-chat threads, branching conversations, agents with custom instructions, context profiles, prompt libraries, and attached reference material. Vision-capable models can work with screenshots, PDFs, code, and other documents. Advanced generation controls expose settings such as temperature, token limits, top-p, top-k, penalties, and system instructions.
- Open BoltAI over another app with a global shortcut.
- Capture a screen region and send it to a model for explanation or analysis.
- Connect local and remote MCP servers, with tools enabled per agent.
- Run local models through Ollama or LM Studio.
- Reuse chats, agents, and projects across Mac, iPhone, and iPad through the product's sync features.
Privacy and ownership
BoltAI states that chats are stored locally and are not used for training. API keys can be encrypted with a user passphrase. This makes it especially relevant to people who want a structured AI workspace while retaining control of providers, credentials, and local conversation data.
