Shinkai

Shinkai lets people build AI agents to automate tasks without writing code. Its desktop app can run agents entirely offline with local models, including workflows that handle sensitive data.

• Agent builder: Create agents with custom instructions, knowledge bases, tools and scheduled tasks through a guided interface that generates code you can also edit.
• Model choice: Use local models such as Llama and Gemma, connect to OpenAI or Anthropic, or access decentralized AI networks from the same system.
• Collaborating agents: Run agents that share context and coordinate work across multiple steps, with persistent memory for ongoing workflows.
• Tools and MCP: Give agents access to browsers, files and external APIs, and connect agent services to Claude or Cursor through Model Context Protocol (MCP).
• Payments and marketplace: Agents can pay providers directly through x402, handle crypto transactions, and be shared or sold through the built-in marketplace.

The desktop app runs on Windows, macOS and Linux. Local models support fully offline processing; choosing a cloud provider sends requests to that external service. Shinkai stores data locally in its Node and supports explicit sharing. A separate web app provides browser access to multiple AI providers and research tools.

Shinkai is free and open source under Apache 2.0. The desktop app lists 4 GB RAM as its minimum and recommends 8 GB.