Guide · Open source
How to use Auto-claude-code-research-in-sleep (ARIS) with Claude Code or Codex
Updated · written with AI from the project's own documentation and checked against our catalogue data
Auto-claude-code-research-in-sleep, usually called ARIS, is a set of plain Markdown skill files for AI coding agents such as Claude Code and Codex. It is aimed at ML researchers and developers who already use one of these tools and want a structured workflow for idea discovery, experiment automation and review loops. After this guide you will know what ARIS needs, how to get the repository, and how the executor-and-reviewer pattern works.
ARISwanshuiyin/Auto-claude-code-research-in-sleep · ★ 18.5k · MIT Install options & downloads Quick start
- 1Confirm your agent. Make sure Claude Code or the Codex CLI is installed and working on your machine.
- 2Clone the repository. Run the git clone command from the README to download the ARIS skill files.
- 3Open the Quick Start section. Follow the Quick Start steps in the README for registering skills with your agent, including the codex reviewer step if it applies to you.
- 4Pick a research task. Choose one of the workflows the README describes, such as idea discovery, experiment automation or a review loop.
- 5Let the reviewer check output. Run the executor and review loop so an independent model checks the executor's work.
- 6Verify results yourself. Read the outputs critically, since the README describes an agent-driven workflow that still needs oversight.
What ARIS is and what it is not
ARIS describes itself as a methodology rather than a platform. It does not add a separate framework or application; the skills run inside an agent you already have installed.
The README says ARIS works with Claude Code, Codex, OpenClaw, or any LLM agent, and lists adaptations for Cursor, Trae, Antigravity, GitHub Copilot CLI and DeepSeek Harness.
- Licence: MIT
- Main language listed: Python
- The README centers on skills and an optional CLI, not a graphical interface
Prerequisites: an agent such as Claude Code or Codex
The skills run inside a supported agent. According to the README, the supported agents include Claude Code and the Codex CLI, so you need one of these installed before ARIS is useful.
The README also notes that the default reviewer is Codex MCP, and that alternative model combinations, including ModelScope-hosted models, do not require a Claude or OpenAI API.
Get the repository
Clone the repository from GitHub using the command shown in the README. This copies the skill files to your machine so your agent can read them.
The README's Quick Start section lists further steps for registering the skills and the codex reviewer. Follow those steps in the repository, since the exact sequence can change between releases.
git clone https://github.com/wanshuiyin/Auto-claude-code-research-in-sleepUnderstand the executor and reviewer pattern
ARIS is built around two roles. The executor drives the work, and an independent model reviews it. The README states that the executor drives and an independent model (Codex MCP by default) reviews.
This split is the core idea: one model checks the output of another, which is useful if you want a second opinion on experiment results or paper claims.
- Executor: the agent that carries out the research task
- Reviewer: a different model that checks the executor's output
- Review loops repeat this check as part of the workflow
Know the limits before you rely on it
ARIS is an agent-driven workflow that still needs human oversight. The README does not promise guaranteed research results, so treat outputs as material to verify, not finished findings.
The README also notes that codex-cli 0.154.0 removed the codex mcp-server entry point that ARIS reviewer calls used, and that ARIS now provides its own bridge over codex exec. If you already installed ARIS, the README points to a re-registration step in its Quick Start section.
- Requires an existing agent; the skills do not run on their own
- Human review of results is still needed
- Check the README's changelog notes if your codex-cli version is recent
What people use it for
- Running an idea discovery pass on an ML research question inside Claude Code
- Automating experiment runs where a second model reviews the executor's results
- Pairing two different models so one reviews the other's paper or report drafts
- Using Markdown skill files as a lightweight alternative to installing a research framework
Frequently asked questions
Is Auto-claude-code-research-in-sleep free?
The repository is licensed under MIT, according to the README. The source does not list prices for any hosted services, so check the repository for any costs that apply to the models or tools you connect.
What do I need to install before ARIS?
The README lists Claude Code or the Codex CLI as the supported agents, so you need at least one of them. The skills run inside those tools rather than as a standalone app.
How do I install ARIS?
Clone the repository with git clone https://github.com/wanshuiyin/Auto-claude-code-research-in-sleep, then follow the Quick Start steps in the README to register the skills with your agent.
Does ARIS need a Claude or OpenAI API key?
The README says alternative model combinations, including ModelScope-hosted models, need no Claude or OpenAI API. The default reviewer is Codex MCP, so check the README for the setup you choose.
Why did my Codex reviener stop working after an update?
The README says codex-cli 0.154.0 removed the codex mcp-server command that ARIS reviewer calls used. ARIS now runs its own bridge over codex exec, and the README points existing users to a re-registration step in Quick Start.
How does ARIS compare to other tools for autonomous research?
The source does not name comparable projects, so it cannot make a direct comparison. The README describes ARIS as a Markdown skill set rather than a graphical app or platform, which is the most useful point to weigh against other options.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
ARIS suits ML researchers and developers who already use Claude Code or Codex and want a structured, plain-Markdown research workflow with a second model reviewing the first. It is a poor fit if you do not have a compatible agent, want a fully hands-off system with guaranteed results, or need a graphical application.
Install options & downloads