About Wisp Science

What it is

Wisp Science is not a new foundation model. It uses the model service configured by the researcher and keeps conversations, files, code, runtimes, tool calls, and generated results in the same project record.

It is designed for literature research, data analysis, scientific database retrieval, and reproducible deliverables. Researchers remain responsible for reviewing methods, statistical decisions, and final conclusions.

Where data goes

By default, project files, sessions, artifacts, and settings stay on the user’s device. Prompts and model responses pass through the selected model provider; calls to a database, remote MCP, agent, SSH/WSL host, or other compute service send the corresponding request to that service.

Project sync must be configured and started manually. Content is encrypted on the device before it reaches a self-hosted relay or a folder managed by a desktop sync client such as Baidu Netdisk or Nutstore; synchronization does not run automatically in the background.

When Feishu/Lark or WeChat remote conversations are enabled, user messages, answer text, and coarse progress updates pass through the bound messaging service. In v0.15.0, Feishu CardKit does not copy model reasoning, tool output, or command output. Fees, quotas, and data policies are governed by the user and each relevant provider.

Relationship to WispTerm

Wisp Science and WispTerm use separate websites, source repositories, and release cycles. The Wisp Science site focuses on scientific workflows, guides, release notes, and installers; WispTerm remains a separate product entry point.