Simplest Agent — AI skill for Claude Code
A simplest AI agent framework.
How to install Simplest Agent
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open Wonderfon/Simplest-Agent and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Simplest Agent does
A simplest AI agent framework. Drop this into Claude Project and you can generate your agent with one sentence prompt
Alternatives in AI
- Agency Orchestrator — 🚀 One sentence → multi-AI-role collaboration → complete plan in minutes 453 ★
- AI Marketing Skills — The marketing fundamentals that make AI actually make you money, written in my own voice, as a ready-to-instal 226 ★
- Semble Rs — Fast, AI-agent-native code search in Rust — hybrid BM25 + semantic, Tree-sitter AST chunking, dependency & imp 210 ★
README
SimplestAgent - Minimalist AI Agent Framework
A lightweight, configurable state machine for building LLM-powered agents with minimal code and dependencies.
"Generate Your Own AI Agent with just a Single Sentence"
This is not just clickbait. It’s possible as at its core, an agent really is this simple.
A loop, a state machine, and some external preset functions. That’s the entire structure of our framework. If you’ve ever called an LLM API, you can easily understand what our source code does.
**Quick Start Guide**:
- Drop
agent.py,agent_config.toml, andrun_agent.pyinto Claude Project knowledge. - For better results, also upload
SimplestAgent_documentation.mdand paste the content inProject Instructions.txtinto Claude's project instructions. - Go to the chat window and tell Claude what agent you want to build.
- Download the generated Python files, follow the AI’s instructions, and run it—your agent is now up and running.
We’ve tested this—Claude Project can generate an AI agent with just one sentence. It should also work if you upload the files to Cursor and call Claude, or use Cline/Roo-Cline.
(However, ChatGPT Canvas won’t work!)
🌟 Key Features
- TOML-based configuration: Define states, prompts, and transitions declaratively
- Single-loop architecture: One LLM call per iteration
- Action dispatching: Register custom functions the LLM can invoke
- OpenRouter API integration: Use any model from OpenRouter's marketplace
- Multi-model support: Use different models for different agent states
- Dev mode: Built-in logging for debugging and development
🚀 Quick Start
Installation
# Install dependencies
pip install openai python-dotenv requests tomli tomli-w
Setup
Create a `.env` file:
OPENROUTER_API_KEY=your_api_key_here
Basic Usage
from dotenv import load_dotenv
import os
from agent
Related Skills
Essence
Compile a Wiki Concept/Guide page into an in-page ## Essence Card — the page's one-sentence model, what it let
Karpathy Wiki Protocol
Battle-tested protocol for AI agents (Claude Code, Codex, Cursor) to build and maintain Karpathy-style wikis.
Art Prompt Framework
🚀 Stop prompting like it's 2023! ART = Act as · Request · Terms. Turn messy asks into prompts so sharp the AI
Planning Doc Rules
Ten rules for writing a planning document — drop-in CLAUDE.md / AGENTS.md system prompt that turns any AI agen
Secretctl
The simplest AI-ready secrets manager. Local-first, single-binary CLI & Desktop app with MCP integration. Neve
Ppt Agent Skill
🎨 World-class AI presentation generator · 26 styles · 18 charts · benchmarked against Linear/Anthropic/Stripe
Related Agents
Coder Gamma
Coder Gamma — minimalist R coder. Prioritises the simplest, shortest correct solution with the least abstracti
Simplicity Critic
Adversarial review of implementation plans and designs for over-engineering. Attacks scope, abstraction count,
Harnie Builder
Pragmatic senior engineer. Implements designs and task instructions with the simplest robust code, then verifi