orx CLI Subcommands: Complete Guide to OpenResearch Command-Line Tools

The orx command-line interface exposes over 20 user-facing subcommands—including login, exp, compute, discover, and up—for managing authentication, experiments, GPU instances, and literature retrieval, with all commands defined in the Command enum in src/main.rs and implemented in the src/commands/ directory.

The alphaXiv OpenResearch repository provides the orx CLI as a Rust-based tool for orchestrating research workflows. Whether you are provisioning GPU compute, tracking experiment runs, or integrating with AI coding agents, mastering the orx CLI subcommands allows you to interact with the platform efficiently from your terminal.

Authentication and Identity

Managing access to the OpenResearch platform starts with authentication commands that handle OAuth tokens and credentials.

orx login and orx logout

The login subcommand initiates a Loopback OAuth flow that opens your default browser for authentication and stores the resulting personal access token locally. According to the source code, this command is defined in the Command enum in [src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L63-L66), with the browser interaction handled by the utility module in src/browser.rs. The logout command ([src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L66-L68)) removes the stored token from your local configuration.


# Authenticate and store credentials

orx login

# Remove stored authentication

orx logout

Project and Organization Management

Organize your research work using commands that interface with local project storage and organizational billing contexts.

orx projects and orx orgs

The projects subcommand lists locally stored projects with optional JSON output formatting, implemented in [src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L69-L71). Similarly, orgs lists organizations available for billing compute resources ([src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L72-L74)).

orx project

For operations on specific projects, the project subcommand accepts nested actions such as view and edit to inspect or modify project metadata ([src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L75-L78)).


# List all local projects

orx projects

# View a specific project

orx project view <project-id>

# List billable organizations

orx orgs

Experiment and Run Lifecycle

The core workflow commands enable creating, monitoring, and managing experimental runs on remote compute.

orx exp

The exp subcommand provides comprehensive experiment manipulation capabilities including status, desc, run, cancel, wake, and wait actions. This is defined in [src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L99-L104) and implemented in the corresponding command module.

orx runs, orx logs, and orx create-experiment


# Create a new experiment node

orx create-experiment <project-id> --title "Baseline Model"

# Check experiment status

orx exp status <exp-id>

# Launch a run on H100 GPUs

orx exp run <exp-id> --backend openresearch --flavor h100_sxm

# Stream logs from a running job

orx logs <run-id> --tail

Compute and Instance Management

Provision and manage standalone GPU resources independent of experiment workflows.

orx compute

Displays the available GPU compute catalog with filtering options for GPU type, provider, and device count ([src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L91-L94)).

orx instance

Manages standalone compute instances with sub-actions create, list, and delete for provisioning resources within specific organizations ([src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L95-L98)).


# Browse available GPU configurations

orx compute

# Provision an H100 instance

orx instance create <org-id> --gpu H100_SXM

# List active instances

orx instance list <org-id>

SSH Key Management

Configure secure access to provisioned instances using the ssh-key command ([src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L97-L100)), which supports add for registering new public keys and list for viewing existing credentials.


# Register a new SSH public key

orx ssh-key add ~/.ssh/id_ed25519.pub

# List registered keys

orx ssh-key list

Agent Integration and Skills

Integrate the orx CLI with AI coding agents through skill management commands.

orx agent, orx skill, and orx install-skills


# Install orx skills for Claude

orx install-skills --agent claude

# Spawn an agent session

orx agent

Library Management

Manage reusable templates and skills in your local library using the skills and templates subcommands ([src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L107-L112)), which allow you to add reusable components to your research workflow.

Literature Retrieval

Access academic papers and search across scientific literature directly from the command line.

orx discover

Executes retrieval primitives including keyword, embedding, openalex, and biorxiv search methods ([src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L117-L121)).

orx paper

Fetches papers by various identifiers including arXiv/alphaXiv IDs, DOIs, or OpenAlex IDs, with optional full-text retrieval ([src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L124-L128)).


# Search for papers by keyword

orx discover keyword "deep learning" --limit 5

# Fetch a specific arXiv paper

orx paper 2101.00001v1

Tooling and Maintenance

Administrative commands for version management, local server operation, and system maintenance.

Version and Update Management

Local Development Server

The serve subcommand starts a local HTTP/SSE daemon on port 4791 (by default), exposing the run store for use by the web UI and coding agents ([src/main.rs](https://github.com/alphaXiv/OpenResearch/blob/main/src/main.rs#L136-L138)).

System Operations


# Check version and available updates

orx version --check

# Update to latest release

orx update

# Launch the local dashboard

orx up

# Start the local API server

orx serve

Summary

The orx CLI provides comprehensive functionality for managing the entire OpenResearch workflow:

  • Authentication: login and logout commands secure your session with OAuth tokens stored in src/config.rs
  • Project Management: projects, orgs, and project commands organize your research workspace
  • Experiment Control: exp, runs, logs, and create-experiment handle the full lifecycle of experimental runs
  • Compute Resources: compute and instance manage GPU catalogs and standalone instances
  • Literature Access: discover and paper enable command-line search and retrieval of academic papers
  • Integration: agent, skill, and install-skills connect with AI coding assistants
  • Maintenance: version, update, serve, and up keep the tooling current and accessible

All commands route through the Command enum defined in src/main.rs (lines 57-110) with implementations residing in the src/commands/ directory.

Frequently Asked Questions

What is the default command when running orx without arguments?

Running orx without subcommands defaults to the up command, which launches the local autoresearch dashboard on http://127.0.0.1:4791. This behavior ensures immediate access to the GUI when users simply type orx in their terminal.

How does the orx CLI handle authentication tokens?

The CLI stores personal access tokens obtained through the login command using the configuration module in src/config.rs. The login subcommand utilizes src/browser.rs to open the system default browser for Loopback OAuth, while logout securely removes these credentials from local storage.

Can I use orx to provision compute without creating an experiment?

Yes, the instance subcommand allows direct provisioning of standalone compute resources using orx instance create <org-id> --gpu <type>, independent of the experiment workflow managed by the exp command. This is useful for ad-hoc development or debugging sessions requiring GPU access.

Where are the command implementations located in the source code?

Each subcommand is defined in the Command enum within src/main.rs (lines 57-110) and implemented in individual modules under src/commands/. For example, authentication logic resides in src/commands/login.rs, while compute management is handled in src/commands/instance.rs, following the dispatch pattern documented in src/commands/mod.rs.

Have a question about this repo?

These articles cover the highlights, but your codebase questions are specific. Give your agent direct access to the source. Share this with your agent to get started:

Share the following with your agent to get started:
curl -s "https://instagit.com/install.md"

Works with
Claude Codex Cursor VS Code OpenClaw Any MCP Client

Maintain an open-source project? Get it listed too →