Tiny agentic loop with Docker sandbox

Name: Tiny agentic loop with Docker sandbox
Availability: InStock
Author: everlier

by everlier·May 13, 2026·1 point·0 comments

Visit Project View on HN

AI Analysis

●●SolidCozyBig Brain

Thirty-line agent loop with Docker sandboxing contains blast radius safely.

Strengths

•Recursive delegation via 'goal' tool iterates subagents until bash checks pass.
•Hot-loaded modular tools allow dropping .mjs files to extend capabilities instantly.
•Works with any OpenAI-compatible API including local models via Ollama or LM Studio.

Weaknesses

•Thirty-line constraint limits sophisticated error handling or complex state management.
•Docker sandbox adds latency; not suitable for rapid iterative development cycles.

Post Description

New in mi v1.7.5: "--sandbox" runs the agent inside Docker with your current directory mounted, to contain the blast radious in case of unexpected attacks.

mi Works with any OpenAI-compatible model (including local ones). Has REPL, skills, and recursive agents and only 30 lines of code, optimised for local models.

--- mi --sandbox -p 'fix tests until all are passing' ---