ARIS vs GenericAgent
Which should you use? A side-by-side look at what each one does, who it suits, what it costs, how hard it is to set up, and where each falls short.
Open source · MIT
Markdown-based skill files that let an AI agent run ML research loops, with a second model reviewing the work.
Open source · MIT
A small, self-improving AI agent that controls your computer and saves successful steps as reusable skills.
The short answer
Choose ARIS if you need…
- ML researchers who already use Claude Code or Codex and want a structured research workflow
- People who want to pair two different models so one reviews the other's work
- Developers who prefer plain Markdown configuration over installing a framework
- Teams experimenting with agent-driven paper review and experiment loops
Choose GenericAgent if you need…
- Developers who want a small, readable agent codebase they can modify
- People experimenting with LLM-driven browser or desktop automation
- Users who want an agent to remember and reuse task workflows over time
- Tinkerers comfortable running Python and supplying their own LLM API key
Side-by-side
| ARIS | GenericAgent | |
|---|---|---|
| Type | AI model / agent | AI model / agent |
| Price | Free and open source | Free and open source |
| Licence | MIT | MIT |
| GitHub stars | 18.5k | 14.4k |
| Setup difficulty | Intermediate | Intermediate |
| How you get it | See the README | Clone and build from source |
| Main language | Python | Python |
| Last code update | 5 days ago | 4 days ago |
What each one is
ARIS
ARIS is a set of plain Markdown skill files for AI coding agents such as Claude Code or Codex. The skills guide an agent through autonomous machine learning research tasks like idea discovery, experiment automation, and review loops, where an independent model checks the executor's output. The README describes it as a methodology rather than a platform, so it does not add a separate framework.
GenericAgent
GenericAgent is an open-source Python framework that lets a large language model control a local computer through a small set of basic tools, including a browser, terminal, filesystem, keyboard and mouse, and screen reading. Each time it solves a new task, it stores the steps as a reusable skill, so its library of skills grows with use. The project says its core is about 3K lines of code and that it uses fewer tokens than many other agents, though the README's efficiency claims are not independently verified here.
Best for
ARIS
- ML researchers who already use Claude Code or Codex and want a structured research workflow
- People who want to pair two different models so one reviews the other's work
- Developers who prefer plain Markdown configuration over installing a framework
- Teams experimenting with agent-driven paper review and experiment loops
GenericAgent
- Developers who want a small, readable agent codebase they can modify
- People experimenting with LLM-driven browser or desktop automation
- Users who want an agent to remember and reuse task workflows over time
- Tinkerers comfortable running Python and supplying their own LLM API key
Strengths
ARIS
- Markdown-only skills with no framework to learn, per the description
- Works across several agents, including Claude Code, Codex, Cursor, and OpenClaw, according to the README
- MIT licensed and actively maintained, with a recent push and frequent releases
GenericAgent
- Core design is small (about 3K lines of seed code and a roughly 100-line agent loop), which makes it easier to read
- Supports several LLM providers named in the README, including Claude, Gemini, Kimi and MiniMax
- Dependencies are described as deliberately light, with no Playwright or LangChain required
Not ideal for
ARIS
- People without access to Claude Code, Codex, or another compatible agent, since the skills run inside those tools
- Users who need a fully hands-off system with guaranteed research results, because the README describes an agent-driven workflow that still requires oversight
- Readers looking for a general-purpose app with a graphical interface; the README centers on skills and an optional CLI
GenericAgent
- Non-technical users expecting a polished one-click app, since setup needs Python and an API key
- Anyone needing strong guarantees about what an agent does on their machine, since it has broad system control
- Users on Python 3.14, which the README says is incompatible with some dependencies
Trade-offs
ARIS
- The README notes a recent codex-cli change that broke the reviewer bridge, so setup may need a re-registration step
- Much of the documentation is in Chinese and English mixed, and the README is long and truncated here, so details are hard to verify quickly
- Results depend on the underlying models and API access, which the README does not fully specify
GenericAgent
- Broad system control means it can take real actions on your computer, so it needs careful supervision
- The README's performance and token-savings figures come from the authors and should be checked for your own use
- Windows terminal rendering can be unreliable, and the README recommends Git Bash
Getting started
ARIS
Clone the repository
git clone https://github.com/wanshuiyin/Auto-claude-code-research-in-sleepGenericAgent
Clone the repository and enter it
git clone https://github.com/lsdefine/GenericAgent.git && cd GenericAgentCreate a virtual environment and install with the UI extras
uv venv && uv pip install -e ".[ui]"Frequently asked questions
What is the difference between ARIS and GenericAgent?
ARIS: Markdown-based skill files that let an AI agent run ML research loops, with a second model reviewing the work. GenericAgent: A small, self-improving AI agent that controls your computer and saves successful steps as reusable skills. ARIS suits ml researchers who already use claude code or codex and want a structured research workflow, while GenericAgent suits developers who want a small, readable agent codebase they can modify.
Are ARIS and GenericAgent free?
ARIS: Free and open source. GenericAgent: Free and open source.
Is ARIS or GenericAgent more popular?
ARIS has more GitHub stars (18.5k) than GenericAgent (14.4k). Popularity isn't the same as fit — check which one matches your use case above.
What are other alternatives to ARIS and GenericAgent?
Other options people compare them with include Hermes Agent, Humanizer, MCP for Blender, Ponytail, last30days.
Keep comparing
Built from our catalogue entries for both products: open-source details come from each project's README (AI-assisted, refreshed weekly) and GitHub data; AI-tool details are editorial. Spotted something wrong? Let us know from either product's page.