Open Source
Explore the latest AI open-source projects from GitHub and HuggingFace.
Explore the latest AI open-source projects from GitHub and HuggingFace.
OpenResearch is a local-first workspace from alphaXiv that turns an existing coding agent, Claude Code, Codex, OpenCode, or Cursor, into a research agent that can review literature, develop hypotheses, run experiments, and produce research artifacts, without bundling a model or agent of its own. The CLI installs on macOS or Linux with a shell script and launches with orx up, which opens a local dashboard at http://127.0.0.1:4791; a beta Windows build is also available but needs Git for Windows, since orx relies on the bash and coreutils it provides. Each research direction can run in its own agent session with an isolated git worktree, while a git-native experiment tree tracks variants and gives every run an immutable archive of its recorded commit, keeping logs, diffs, files, results, and artifacts tied to the run that produced them. In autoresearch mode, OpenResearch can run the loop unattended: propose an idea, change code, launch an experiment, inspect the evidence, and decide what to try next, with agents exploring directions in parallel while the experiment tree preserves their lineage. The same committed snapshot can run locally, over SSH, or on Slurm, Kubernetes, Ray, Hugging Face Jobs, Modal, Tinker, or managed OpenResearch compute; publishing the repository is not required. orx up --remote user@host runs the workspace next to remote GPUs from a local browser, though the remote service binds to loopback with no application-level authentication, so other users on that host can reach it. OpenCode can also point at a locally hosted model through LM Studio, oMLX, Ollama, or another OpenAI-compatible endpoint instead of a cloud provider. OpenResearch runs on 127.0.0.1 with a local SQLite store, and creating a project or launching a run does not publish the underlying code; an openresearch.sh account is only used for service-owned capabilities such as organizations and managed compute. Official release builds send opt-out, coarse usage analytics tied to a random installation ID, excluding code, prompts, file contents or paths, repository names, and project or experiment identifiers, and this can be disabled with orx telemetry off; source and dev builds send none. The project is written mostly in Rust and TypeScript and is MIT licensed.