ScienceClaw

ScienceClaw is a personal research assistant for scientific and academic workflows. It is built with LangChain DeepAgents and AIO Sandbox infrastructure, and the product text says it was redesigned beyond OpenClaw with an emphasis on security, transparency, and ease of use.

The system is described as running entirely inside Docker containers. The agent cannot access the host system, personal files, or environment variables, and code runs in an isolated sandbox. Its workflow is presented as fully traceable, with each step visible from web search and data crawling through reasoning, tool invocation, and final report generation. ScienceClaw also includes a skill system that uses structured instruction documents to guide the agent through complex multi-step workflows.

Its toolset is described as extensive, with 1,900+ built-in scientific tools across eight domains. The listed domains are drug discovery and biomedicine, astronomy and space science, earth and environmental science, chemistry and materials science, biodiversity and ecology, social science and statistics, academic literature, and data science and computing. The page also names built-in web search and web crawling, ToolUniverse integration, sandbox tools for file operations and code execution, and a custom @tool mechanism that supports user-defined Python functions and hot-loading. Content generation is said to cover four formats: PDF, DOCX, PPTX, and XLSX, including PDF reading and creation, Word documents, PowerPoint slide decks, and Excel spreadsheets with CSV/TSV data processing.

Delivery is local: the page states 100% local deployment and gives setup instructions for Windows, macOS, and Linux. It can be launched with a single command, and the Windows instructions mention a desktop shortcut after installation.