We build autonomous research systems that take a project from a one-line topic to running experiments, with no human-written experimental code. Agents plan, implement, audit and review each other in closed loops, and every handoff goes through a file on disk — so a run is recoverable, auditable, and reusable across projects. The workflow stays minimal and explicit: topic → idea → proposal → experiment.
Everything here is built on Prompt Economy: treat prompt engineering as engineering, and minimize the engineering effort it demands from humans.
| Agon | The core system. A Claude Code plugin for autonomous AI research, deployed across more than ten research domains. |
| agon-artifacts | Example data workspace for Agon — topics, ideas, proposals, and experiment workspaces. |
| AgonSR | Automated symbolic regression: a Claude Code plugin for symbolic ansatz search. |
| AgonAlpha | Agent-based framework for automated alpha discovery. |
| AgonReproduce | Prompt-first auditing of research claims: literature investigation, direct reproduction, and independent review in one traceable workflow. |
Clone the plugin and an artifacts workspace side by side, then run Claude Code from the workspace:
git clone https://github.com/AutoResearch-Factory/Agon.git
git clone https://github.com/AutoResearch-Factory/agon-artifacts.git
cd agon-artifacts
claude --plugin-dir ../Agon --dangerously-skip-permissions --model claude-sonnet-5[1m]
Then drive the research forward with /idea-tick, /proposal-tick, /experiment-tick, and /deep-lit-tick. See the Agon README for details.
Agon: An Autonomous Large-Scale Omnidisciplinary Research System Built on Prompt Economy
Youran Sun, Xingyu Ren, Chugang Yi, Jiaxuan Guo, Kejia Zhang, Jianda Du, Haizhao Yang
arXiv · Project page