A minimal AI agent built from scratch — no agent framework, just Python, the OpenAI SDK, and a `while` loop. Accompanies the Medium article "Build Your Own AI Agent from Scratch." Three execution modes: `local` (Ollama only), `remote` (cloud provider), `mixed` (local orchestrates, delegates to remote).
6 files for Claude Code, Codex, OpenCode and Gemini CLI: deploy, running-the-agent, visual-check, mini_agent AGENTS.md and 2 more — 522 tokens loaded in every session.
AGENTS.md A 236 tok CLAUDE.md A 242 tok GEMINI.md A 44 tok .claude/skills/deploy/SKILL.md A 50 tok .claude/skills/running-the-agent/SKILL.md A 55 tok .claude/skills/visual-check/SKILL.md A 53 tok These files are sergenes/mini_agent's own configuration — they tell Claude Code, Codex, OpenCode and Gemini CLI how to work on this repository, so they are not mods to install elsewhere. Copy one as a starting point and replace the parts that are about this project.