Jan

Jan packages open-source AI into a cross-platform product that runs locally or connects to remote models. It supports custom assistants, agentic capabilities via MCP servers, a model hub, and a local OpenAI-compatible API server for self-hosted deployments. Jan targets researchers, developers, and power users who want control over models, connectors to apps and services, and an extensible platform built in public.

• Local model execution — Run open-source models locally; install Jan Desktop and add models through the Jan Hub or local model manager.
• Remote model connectors — Plug in cloud providers and remote models by adding API keys in settings to use OpenAI, Anthropic, Cohere, Gemini, Mistral, Hugging Face, and others.
• Custom assistants — Create and configure assistants from the Assistants interface to tailor behavior and workflows for specific tasks.
• MCP / agentic capabilities — Enable MCP servers to give models tool use and agent-like behavior; deploy a local MCP server or connect to an existing one.
• Local OpenAI-compatible API server — Expose a local API compatible with OpenAI endpoints to integrate Jan into existing tooling or self-hosted services.
• Connectors and integrations — Link email, files, notes, search, and productivity apps (Gmail, Notion, Slack, Google Drive, Jira, Figma, YouTube, etc.) through the Connectors settings.
• Cross-platform desktop and web — Use Jan Desktop on macOS, Windows, and Linux or the Jan Web interface for browser access; install platform-specific builds from the downloads or run the server components for headless deployments.