Agent Army

100+ production-grade AI agents on one tiny, model-agnostic core.
Every agent runs on Claude, GPT, Gemini, DeepSeek, Groq, OpenRouter or a local Ollama model, without code changes.
One command to run any of them. Eval cases in every agent. Zero dependencies.

Quick start · All agents · How it works · Course · Add an agent

---

Why Agent Army

Most "awesome AI apps" collections are a pile of demos: every folder uses a different framework, a different UI and a different provider, has its own requirements file, and comes with no way to know if it still works. You clone 100 projects and learn 100 setups.

Agent Army is one system:

	Typical demo collections	Agent Army
Setup	A new requirements.txt per project	pip install -e . once, zero third-party dependencies
Switching models	Rewrite the code	`--model ollama