Open Science Desktop
An AI research workbench runs on macOS, Windows, Linux or a self-hosted server, with OpenCode model support and traceable research outputs.
Alternative to:

Open Science Desktop brings research agents, notebooks and project files into one workbench for researchers who want to inspect the work behind an AI-generated result. Agents can carry a research topic through literature review, experiments, analysis and manuscript writing.
- Research artifacts: Specialist agents produce literature surveys, experiment code, figures and papers as files you can inspect and revise.
- Traceable results: Figures, tables, notebooks and reports link to the code, inputs, environment, model output and conversation that produced them.
- Experiment records: Local execution, SSH/Slurm jobs, Modal and notebook batches retain reproducible run records; remote jobs can be submitted, tracked and canceled from the app.
- Model and agent choice: The workbench supports models available through OpenCode, or compatible Agent Client Protocol agents such as Codex, Gemini CLI and Claude Code.
- Persistent projects: Project and global memory carry context between sessions, while split panes let researchers work with different models side by side.
It runs on macOS, Windows and Linux, or as a self-hosted server with browser access. Sessions, data, notebooks and run records live in local folders by default. Model-provider access depends on the host's credentials, and remote compute sends work to the selected service. The optional authenticated browser gateway keeps API keys on the host and provides read-only or full access.