wanshuiyin/SuperARIS

What it does

SuperARIS is a command-line tool that turns several coding-agent CLIs into a persistent research team. One main agent owns the research direction and delegates work; an engineer role runs experiments and checks evidence; a research-and-writing role develops arguments and drafts paper sections. Each role keeps its own session, model choice and client (Codex, Claude Code or Antigravity), so you can join a specific teammate's conversation and continue later.

The orchestration layer is built on ARIS, whose skills cover the research pipeline, idea discovery, experiment bridging, review loops and paper writing. SuperARIS registers tasks, result files and reports so the main agent can collect finished work and decide the next authorized step even after the workbench window is closed. Project goals, decisions and results stay in inspectable project notes, which later sessions read back when resuming project context.

Quick start

Prerequisites: Python 3.10+, Git and pipx, plus the native CLIs you plan to use (codex, claude, agy) with model access already configured. Herdr is the default workbench; Orca is optional.

pipx install "git+https://github.com/wanshuiyin/SuperARIS.git"
superaris

In the wizard, pick the interface language, project, workbench, team and models, then confirm the workspace. Presets include an all-Codex team, Claude Code main/writing with Codex engineering, and Claude Code / Codex / Antigravity. To skip the wizard for an existing project directory, start one workbench explicitly:

superaris start . --host herdr
# or, after installing the official Orca app:
superaris start . --host orca

A custom team can assign clients per role, for example:

superaris start /path/to/project --role main=agy --role engineer=claude --role writer=agy

After startup, give the main agent a task or open the team and discuss next steps. Reopening an existing project keeps its recorded workbench and team. Windows users run the same pipx install inside Windows Terminal or PowerShell after py -m pipx ensurepath; see the Windows setup guide. Installation alternatives, including a local wheel and an existing ARIS checkout, are documented in the dependency notes.

Usage constraints

• Requires Python 3.10+ on macOS, Linux or Windows, and at least one supported agent CLI with configured model access; native CLI login and trust prompts remain interactive.
• Depends on Herdr (installed through its official Windows installer when missing) or a separately installed Orca Terminal v1 for the optional workbench.
• ARIS resources are obtained at a pinned revision when needed; an existing ARIS installation can be selected with SUPERARIS_ARIS_REPO.
• Follow-up behavior differs by client: Codex and Herdr AGY sessions marked as automation-owned can continue automatically when idle, while Claude Code and Orca AGY sessions work through notices.
• The release is labeled Preview v0.1.0, so interfaces and workflows may still change. The project is MIT licensed LICENSE.

HotRepo's take

SuperARIS appeared on HotRepo's new-project list on 2026-11-14, recording 7 new stars over a roughly three-hour window and closing that day's sampling at 108 stars. The codebase is Python and the README is detailed enough to trace concrete features — update checking, per-role model selection, usage inspection and Windows support — back to merged pull requests from the preceding days update PR custom team PR. What stands out is the design choice of keeping long-lived per-role sessions and project notes rather than one-shot prompts, which is also its main cost: it assumes users already have several agent CLIs and are comfortable managing them. The sample is too short to say anything about adoption.