Overview

OpenResearch is a local-first workspace for research agents. It turns the coding agents you already use into research agents that review literature, develop hypotheses, run experiments, and produce research artifacts.

How it works

• Bring your agent. Claude Code, Codex, OpenCode, Cursor, and Google Antigravity work natively, and OpenCode can run local models.
• Ground ideas in the literature. Agents search papers on alphaXiv, OpenAlex, bioRxiv, and PubMed.
• Track every experiment. Each project is a git repository with a tree of experiments. Runs, logs, metrics, and results are recorded locally so agents learn from past results.
• Choose your compute. Run locally, over SSH, on Slurm, Kubernetes, Ray, Modal, Hugging Face Jobs, Tinker, or managed OpenResearch GPUs.
• Own your research. Projects, conversations, experiments, logs, and artifacts stay on your machine.

Quick start

1. Install the desktop app for macOS, Windows, or Linux.
2. Pick a coding agent and sign in to it. The app can install it for you.
3. Create a project from a blank repository, an existing folder, or a paper you want to reproduce.
4. Chat with the agent: ask it to survey related work, propose experiments, and run them.

Windows and Linux builds are on the Install page.

Your OpenResearch account

The app works without an account. Sign in with orx login or from the app to use managed compute, which is billed at cost from a prepaid balance. The web dashboard's Compute and Instances tabs show live GPU prices and your running machines.