TL;DR: We’re launching Hopper, an agentic development environment for IBM mainframes. It combines a real TN3270 terminal and an AI agent that can operate across z/OS workflows like COBOL, VSAM, and CICS.

Download it here: https://www.hypercubic.ai/hopper

Video walkthrough: https://www.youtube.com/watch?v=q81L5DcfBvE

You can also request access and immediately get a mainframe user account to try it.

Problem

Mainframes still run a huge amount of critical infrastructure: banking, payments, insurance, airlines, government programs, logistics, and core enterprise operations.

A lot of that software is written in COBOL and runs on IBM z/OS. These systems are reliable, secure, and deeply embedded into business operations, but the development environment looks nothing like modern software development.

Modern AI coding tools assume repos, files, shells, package managers, test runners, and CI pipelines.

Mainframe development is an entirely different computing paradigm.

A simple COBOL change might require finding the right source member, checking copybooks, locating compile JCL, submitting a job, reading JES/SYSPRINT output, interpreting condition codes, patching fixed-width source, and resubmitting.

Much of this work is structured and repetitive, but today it still depends heavily on expert humans who know how to navigate the terminal, interpret output, and follow local conventions.

A chatbot next to a terminal is not enough.

For AI to be useful here, the agent needs to operate inside the mainframe environment itself.

Solution

Today we’re launching Hopper, an agentic development environment for mainframes.

Hopper combines three things:

1. A real TN3270 terminal
2. Mainframe-aware panels for datasets, members, jobs, and spool output
3. An AI agent that can operate across z/OS workflows

The agent can navigate ISPF, inspect datasets, write and edit COBOL, generate and modify JCL, submit jobs, parse JES spool output, analyze return codes, query VSAM, interact with CICS, and explain failures.

Why this matters

The mainframe is not just a codebase. It is an operating environment with its own workflows, conventions, interfaces, and failure modes.

Most AI coding tools treat legacy systems like static repositories. That misses a huge part of the real work.

In mainframe environments, the important context often lives across terminal screens, datasets, job output, copybooks, CICS regions, VSAM files, scheduler conventions, and decades of operational knowledge.

Hopper’s design principle is simple:

Preserve the fidelity of the mainframe environment, but make it accessible to AI agents.

Once agents can safely operate inside the mainframe, new workflows become possible:

• Faster job debugging
• Automated documentation
• Safer COBOL changes
• Test generation
• Migration planning

We think this is the missing layer between today’s AI coding tools and the mission-critical systems that still run much of the world.

Who we are

We’re Sai and Aayush, former Apple AIML engineers, and we’re building Hypercubic to bring AI-native maintenance and modernization infrastructure to mainframe and COBOL systems.

🤝 What we’re looking for

• Intros: Connections to CIOs, CTOs, VPs of Engineering, or Heads of Modernization at Fortune 500 companies, particularly in banking, insurance, manufacturing and retail.
• Enterprises running COBOL or mainframe workloads exploring modernization or knowledge retention.
• Systems integrators / consultancies seeking AI-powered tooling for mainframe understanding and modernization.