Agent Mini packs a usable local AI agent into about 3,000 lines of Python
Lordrovks · reddit · 2026-07-28
A developer released Agent Mini, a minimal local-first AI agent that is small enough to read and extend without a heavy framework stack.
The project is about 3,000 lines of Python, uses Ollama by default, and includes practical tools for shell, files, web search, memory, and vision. The author explicitly avoids LangChain and LiteLLM, instead building on asyncio, httpx, and a compact ReAct loop tuned for local or smaller models.
The motivation is clarity: the author wanted an agent that is useful in practice without becoming a black box. Installation is simple via PyPI:
- pip install agent-mini
- agent-mini init
- agent-mini chat
The post is looking for feedback from others experimenting with local agents.
More from coding & agent
- Live session promises a from-scratch GRPO walkthrough and TRL experiments — moonsandhues · 2026-07-28
- Claude Opus 5 Generates 3D Neurovascular Simulator With a Single Prompt — BraydonDymm · 2026-07-28
- A PR daemon turns reviewer comments into fix PRs so humans only do the final review — MikkoH · 2026-07-28
- Resetting Claude Context: Developers Debate Memory vs. Clean Slates — mobileraj · 2026-07-28
- SAP’s TRACE preserves tool knowledge and reaches 86% recall with greedy decoding — SAP · 2026-07-28
- Reddit debate asks why coding agents still run planning and review on the same expensive model — Neat_Initiative_7780 · 2026-07-28