# OpenResearch

> A local-first desktop workspace for running parallel research agents using Claude Code, Codex, OpenCode, or Cursor to review literature, run experiments, and produce research artifacts.

OpenResearch is a local-first workspace for research agents and autoresearch, built by alphaXiv and released under the MIT License. It lets researchers spin up multiple AI coding agents — Claude Code, Codex, OpenCode, or Cursor — as independent research agents that work in parallel, each in its own isolated git worktree. The project is available as a desktop app for macOS and Windows (beta) and as a CLI for macOS and Linux.

## What It Is

OpenResearch is an agent harness and experiment management tool designed specifically for the research workflow. Rather than a general-purpose coding assistant, it wraps existing AI agents with the infrastructure needed to run reproducible experiments: parallel sessions, git-native experiment trees, immutable run archives, and evidence tracking. The local dashboard (served at `http://127.0.0.1:4791`) provides a unified view of all running agents, experiments, and their outputs.

## How the Research Agent Workflow Works

The core workflow centers on giving each research direction its own agent session:

- **Parallel exploration**: Each direction gets an independent agent session and isolated git worktree, so multiple hypotheses can be tested simultaneously.
- **Reproducible experiments**: Every run receives an immutable archive of its recorded commit, and variants are tracked in a git-native experiment tree.
- **Evidence in context**: Logs, diffs, files, results, and artifacts stay tied to the work that produced them.
- **Autoresearch loop**: OpenResearch can run the full loop autonomously — propose an idea, change code, launch an experiment, inspect evidence, and decide what to try next.

## Deployment and Compute Flexibility

OpenResearch is designed to run anywhere the researcher's work lives. The same committed source snapshot can execute locally, over SSH, or on Slurm, Kubernetes, Ray, Hugging Face Jobs, Modal, Tinker, and managed OpenResearch compute. A single command (`orx up --remote user@host`) moves the workspace next to remote GPUs while the browser dashboard stays on the local laptop. The project also aggregates compute offerings from multiple providers through an OpenResearch compute marketplace for teams that need to scale beyond local hardware.

## Local Ownership and Privacy Model

All projects, conversations, experiments, runs, logs, code, and artifacts remain on the user's machine by default. 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 required 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 — no code, prompts, file contents, paths, repository names, tokens, emails, or project identifiers are included. Source and development builds send no analytics at all.

## Update: v0.2.1

The latest release is v0.2.1, published on September 12, 2026. The repository was created in June 2026 and has accumulated over 1,600 GitHub stars and 123 forks in its first few months, signaling rapid early adoption in the research community. Windows support is currently in beta and requires Git for Windows. The project is written primarily in Rust and is maintained by alphaXiv, the organization behind the alphaXiv preprint discussion platform.

## Features
- Parallel research agent sessions in isolated git worktrees
- Git-native experiment tree with immutable run archives
- Autoresearch loop: propose, code, experiment, inspect, iterate
- Support for Claude Code, Codex, OpenCode, and Cursor agents
- Local dashboard at 127.0.0.1:4791
- Remote compute via SSH, Slurm, Kubernetes, Ray, Hugging Face Jobs, Modal, Tinker
- Compute marketplace aggregating multiple providers
- Local SQLite store — no code published by default
- CLI with project, run, log, and paper discovery commands
- Agent skill installation via orx install-skills
- Support for local models via LM Studio, oMLX, Ollama, or custom endpoints
- Opt-out usage analytics with no code or prompt data collected

## Integrations
Claude Code, Codex, OpenCode, Cursor, LM Studio, oMLX, Ollama, Slurm, Kubernetes, Ray, Hugging Face Jobs, Modal, Tinker, SSH, Git, arXiv

## Platforms
WINDOWS, MACOS, LINUX, WEB, API, CLI

## Pricing
Open Source

## Version
v0.2.1

## Links
- Website: https://openresearch.sh/
- Documentation: https://openresearch.sh/docs
- Repository: https://github.com/alphaXiv/OpenResearch
- EveryDev.ai: https://www.everydev.ai/tools/openresearch