Hopper AI Agent Launches to Modernize Mainframe Development for Critical Infrastructure

Hypercubic today announced the launch of Hopper, an agentic development environment that brings artificial intelligence directly into the mainframe and COBOL ecosystem. The tool enables AI agents to operate within TN3270 terminal sessions, navigating ISPF panels, datasets, and JES spool output — automating repetitive tasks that have long slowed development on IBM z/OS systems. Download Hopper or request immediate access to a mainframe user account.

“Mainframes still power the backbone of global banking, payments, insurance, and government services,” said Sai, co-founder of Hypercubic. “But their development environment hasn’t changed in decades. Hopper lets AI work inside that environment the way a human expert would — navigating screens, submitting jobs, and interpreting results.” A video walkthrough demonstrates the tool in action.

Background

Mainframes run surprising amounts of critical infrastructure. From banking and payments to airlines, logistics, and government programs, these systems process enormous transaction volumes daily. Much of that software is written in COBOL and runs on IBM z/OS.

The development environment differs starkly from modern cloud or Unix-style workflows. Instead of GitHub, shell commands, and CI pipelines, developers use TN3270 terminal sessions, ISPF panels, partitioned datasets, JCL, JES queues, and spool output. “A simple COBOL change can involve finding the right source member, checking copybooks, locating compile JCL, submitting a job, reading SYSPRINT output, interpreting condition codes, and resubmitting — all through fixed-width screens and function keys,” notes co-founder Aayush.

This work is well-defined and repetitive, making it a strong candidate for agentic AI. However, a chatbot beside a terminal is not enough. The agent must operate inside the mainframe environment. Hopper combines three components: a real TN3270 terminal, mainframe-aware panels for datasets/members/jobs/spool, and an AI agent that can act across those z/OS surfaces.

Debugging Example: A Tiny COBOL Program

The following COBOL code has a typo (CUSTOMER-BALNCE instead of CUSTOMER-BALANCE):

IDENTIFICATION DIVISION.
PROGRAM-ID. PAYCALC.
DATA DIVISION.
WORKING-STORAGE SECTION.
01  CUSTOMER-BALANCE     PIC 9(7)V99.
PROCEDURE DIVISION.
    ADD 100.00 TO CUSTOMER-BALNCE
    DISPLAY "UPDATED BALANCE: " CUSTOMER-BALANCE
    STOP RUN.

A human would submit this via JCL and check output. Hopper can automate the entire cycle — find the source, submit the compile, parse SYSPRINT, identify the error, and even propose a fix. “We built Hopper to handle that kind of pattern recognition and execution loop,” said Sai.

What This Means

Hopper addresses a pressing need: modernizing mainframe development without replacing the systems themselves. Given that mainframes remain deeply embedded in business operations, any AI tool must integrate with existing workflows rather than disrupt them.

“This isn’t about ripping and replacing — it’s about augmenting the expert human,” explained Aayush. “Our goal is to reduce the cognitive load of navigating hundreds of panels and datasets, so developers can focus on logic and business rules.” The launch follows Hypercubic’s earlier Launch HN and represents a step toward making mainframe development more accessible to a new generation of engineers.

Industry observers see potential. “If Hopper can reliably automate even 20% of the routine tasks a COBOL developer faces, it could shorten change cycles significantly,” said Dr. Elena Torres, a senior analyst at TechInsights. “Large financial institutions have been desperate for tools that bridge the gap between legacy systems and AI.” Hypercubic plans to release additional features later this year, including deeper integration with CICS and VSAM.