What OpenResearch does
OpenResearch is a local-first workspace that turns coding agents into research assistants. It works with Claude Code, Codex, OpenCode, Cursor, and Google Antigravity, letting them review literature, develop hypotheses, run experiments, and produce research artifacts. The tool runs on macOS 11+, Windows (beta, requires Git for Windows), and Linux (needs glibc 2.35+).
Installation on macOS or Linux starts with a single command: curl -LsSf https://openresearch.sh/install.sh | sh, followed by orx up, which opens a local dashboard at http://127.0.0.1:4791. On managed Macs, device policies may block the unsigned CLI, so OpenResearch offers a signed and notarized desktop app instead. Windows users get a beta installer that runs without administrator privileges, though the installer is not yet signed.
Built for parallel research agents
Each research direction gets an independent agent session and isolated git worktree, so multiple hypotheses can be explored simultaneously. Experiments are tracked in a git-native tree, and every run receives an immutable archive of its recorded commit. Logs, diffs, files, results, and artifacts stay tied to the work that produced them.
Users choose their agent per session and their compute per run. OpenResearch supports local execution, SSH, Slurm, Kubernetes, Ray, Hugging Face Jobs, Modal, Tinker, and managed OpenResearch compute. A remote workspace can run next to GPUs while the browser stays on a laptop: orx up --remote user@host. The remote service binds to loopback and has no application-level authentication, so other users on that host can reach it.
Autoresearch and the full loop
OpenResearch can run autonomously: propose an idea, change the code, launch an experiment, inspect the evidence, and decide what to try next. Multiple agents can explore different directions in parallel while the experiment tree preserves their lineage. The same committed source snapshot can run across any supported compute environment without publishing the repository.
CLI commands cover the workflow: orx projects, orx runs <project-id>, orx logs <run-id>, orx exp run <experiment-id>, orx discover keyword <query>, and orx paper <arxiv-id-or-doi>. The orx install-skills command adds the OpenResearch skill to supported coding agents.
Local ownership and telemetry
OpenResearch runs on 127.0.0.1 with a local SQLite store. Creating a project or launching a run does not publish code. An openresearch.sh account is only needed for service-owned capabilities such as organizations and managed compute.
Official release builds send opt-out, coarse usage events tied to a random installation ID. They do not include code, prompts, file contents or paths, repository names, tokens, emails, or project and experiment identifiers. Users can disable this with orx telemetry off or pass --no-telemetry to individual commands. Source and development builds send no analytics. The orx feedback command files product feedback, with bug reports including reproduction detail while omitting sensitive information.
Why this matters for researchers and developers
OpenResearch addresses a practical problem in AI-assisted research: keeping experiments reproducible while letting agents work in parallel. The git-native experiment tree means every run has a recorded commit, which matters if you need to audit results or defend a finding. For teams running on institutional infrastructure, the SSH and Slurm support means the same snapshot can move from a laptop to a cluster without code changes.
If you are learning to build research workflows with coding agents, understanding tools like this pairs well with structured training on AI for Scientists Courses and AI Agent Courses. The local-first design also means your literature reviews, hypotheses, and experimental code stay on your own machine unless you choose otherwise - a meaningful distinction for unpublished work.
Your membership also unlocks:
