cbbde3890f
* Add a History view of experiment lineage and run attempts Co-authored-by: lc140159_gsk <211554985+lc140159_gsk@users.noreply.github.com> * Address review: archived filter, overlay focus, fold counts and narrow layout 1. History ignored "Show archived experiments". App now passes the setting, and History hides archived experiments unless it is on while still building lineage from every experiment, so a shown child of an archived parent keeps its "from ..." lead-in. When every experiment in scope is archived, History shows the same hint as the table. 2. In a narrow view the detail pane covered the list but focus stayed on a covered row. The covered list is now inert until the pane closes, and the pane takes focus when it covers the list. Closing it still returns focus to the row. 3. A folded branch counted later same-chapter experiments reached through another chapter, which also render as their own roots. The fold now follows the same in-chapter children the layout draws. 4. Search could fold a side branch holding a match. Nothing folds while a search is active. 5. The From column stayed at compact widths. It now follows the compact rule like Attempts and Latest. 6. Chapter headers kept status counts and dates on one line at compact widths, squeezing the title. The summary is dropped at compact widths. The width rules move into historyFit() in the model so they can be tested without a browser. Co-authored-by: lc140159_gsk <211554985+lc140159_gsk@users.noreply.github.com> * Add Polish translations for the new strings Upstream added Polish as a ninth catalogue (#626), so the strings this change introduces need a Polish entry for the catalogue lint to pass. Co-authored-by: lc140159_gsk <211554985+lc140159_gsk@users.noreply.github.com> * Integrate lineage into experiment table and overview * Trim trailing whitespace in lineage component * Hide experiment lineage by default * Preserve tab gestures in experiment lineage links --------- Co-authored-by: lc140159_gsk <211554985+lc140159_gsk@users.noreply.github.com> Co-authored-by: Daniel Kim <sox8502@gmail.com>
OpenResearch
The local-first harness & workspace for research agents.
an 
alphaXiv project
English · Español · 한국어 · 简体中文 · 日本語
Whether you’re looking for a copilot or an autoresearch tool, OpenResearch will accelerate your work. Launch research agents that can review literature, develop hypotheses, run experiments, and produce research artifacts.
macOS 11+ · Windows beta requires Git for Windows · Linux app needs glibc 2.35+
Get started
We recommend using the OpenResearch desktop app. Download it for your platform:
- macOS
- Windows (Beta)
- Linux: x86_64 · ARM64
Prefer the standalone CLI? Install on macOS or Linux
curl -LsSf https://openresearch.sh/install.sh | sh
orx up
orx up opens the local dashboard at http://127.0.0.1:4791.
On a managed Mac, device-management policies may block the CLI installed by
install.sh because it is not yet signed. Use the signed and notarized desktop
app instead. You can also install orx for your terminal from the app's
Settings → Updates → Install the orx command.
Create an account at openresearch.sh to receive email updates and use managed OpenResearch compute.
How it works
Use your favorite coding agent. OpenResearch works natively with 
Claude Code,
Codex,
OpenCode,
Cursor, and
Google Antigravity.
You can also use a local model with OpenCode through
LM Studio,
oMLX,
Ollama, or a custom endpoint.
Ground research hypotheses in the latest literature. OpenResearch integrates with 
alphaXiv, 
bioRxiv, and 
PubMed so agents can find relevant papers, develop informed hypotheses, and connect their experiments to existing research.
Choose your compute. Run experiments locally or use SSH,
Slurm,
Kubernetes,
Modal,
Tinker,
Ray,
Hugging Face Jobs, and
managed OpenResearch compute.
Give agents a memory of every experiment. OpenResearch records every experiment in a local SQL database and keeps its logs, code, and artifacts on your machine so that agents can learn from past results and decide what to try next. The desktop app makes it easy to see what your agents have done and inspect the evidence behind each result.
Keep and own your research. Run entirely locally. We don't collect your code or agent traces. Your projects, conversations, experiments, logs, and artifacts stay on your machine, under your control.
Usage analytics
Official release builds send opt-out, coarse product analytics events and detailed agent feedback in the background tied to a random installation ID. Your code, prompts, file contents or paths, repository names, tokens, emails, or project and experiment identifiers are not collected. Anonymized token/message counts are used to identify which models are being used across OpenResearch and can be seen at https://openresearch.sh/rankings. Feedback reports from the agent are sent for bug reports, feature requests, and general frustrations about OpenResearch. Agents are instructed to sanitize reports to exclude any sensitive information about your research or code.
You can turn off all usage analytics and agent feedback in the desktop app's
Settings or with orx telemetry off.