Zro

What it does

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.

Who it is for

Zro targets developers and teams using AI coding agents that require privacy and speed. It is for users of popular tools like Claude Code, Codex, Cursor, Cline, and others who want to run open-weight models without data retention or training on their prompts and completions. It also appeals to organizations that need multi-region inference for compliance or latency reasons, especially those in regulated industries where data privacy is paramount.

Why it matters

As coding agents become more prevalent in development workflows, data privacy is a growing concern. Many inference providers retain prompts or use them for model training, creating risks for proprietary code and sensitive information. Zro addresses this by promising zero retention by default and no training on customer data, offering a compelling value proposition for enterprises and privacy-conscious developers. Additionally, by focusing on open-weight models—which are increasingly competitive with closed-source systems—Zro provides a faster and more private alternative for coding tasks without sacrificing model quality.

Launch signal

Zro launched on Product Hunt as a promoted product, with the tagline "Private inference for coding agents." The website shows active integrations with multiple coding tools and includes several positive testimonials praising the service's speed and privacy features. The product offers a $20/month starter plan with $60 of inference spend, and usage packs without subscription. The launch appears focused on the developer community, with a clear emphasis on privacy and performance.

Brand and naming

The name "Zro" is a stylized spelling of "zero," directly communicating the core promise of zero data retention and zero training on customer data. It is short, catchy, and memorable, making it easy for developers to recall. The positioning as "private inference for coding agents" clearly distinguishes Zro from general-purpose inference API providers by targeting a specific high-value use case with a strong privacy guarantee. The branding uses a sleek, minimalist mark that aligns with the technical audience.