Zro
Private inference for coding agents
- Category
- AI & Machine Learning
- Launched
- July 16, 2026
- Pricing
- paid
- Website
- zro.moonmath.ai
Zro is a private inference endpoint designed specifically for coding agents. It serves open weight coding models—currently MiniMax M3, GLM 5.2, and Kimi K2.7 Code, with DeepSeek V4 Pro coming soon—across multi region infrastructure. The service defaults to zero request retention and never trains on customer data. Zro is optimized for long context, multi turn coding sessions, using HyperQuant compression, custom attention kernels, and hardware aware deployment on AMD, NVIDIA, and TPUs. It provides both OpenAI compatible chat completions and Anthropic compatible Messages endpoints, and offers a CLI tool ( @moonmath ai/zro npm package) to launch supported coding agents (Claude Code, Codex, Cursor, Cline, Pi, OpenClaw, OpenCode, Hermes) with temporary session config.
Discovered via
- Product Hunt Today · · Zro: Private inference for coding agents | Product Hunt