> ## Documentation Index
> Fetch the complete documentation index at: https://docs.hypercubic.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Overview

> Hopper is a desktop app for z/OS engineers. It bundles a TN3270 terminal, dataset and job browsers, and an AI agent that drives ISPF, writes JCL, and watches JES on your behalf. Runs on macOS, Windows, and Linux.

Hopper connects to your z/OS system and gives you a modern interface plus an AI agent that reads source from PDSes, submits and watches JCL, decodes VSAM records, and writes changes under your supervision.

Four panels share one mainframe session:

1. **A TN3270 terminal.** A clickable, copy-pasteable rendering of the mainframe's 24×80 green-screen interface.
2. **A datasets browser.** Filter, click, and preview mainframe files inline without typing IDCAMS commands.
3. **A jobs browser.** Watch JES queue entries change status in real time without opening SDSF.
4. **A chat agent.** A panel that runs alongside the terminal. It reads source, submits jobs, decodes records, and writes changes under your supervision.

Hopper connects to your z/OS system over standard TN3270 (and FTP, if you've configured credentials). Nothing runs on the mainframe; everything runs on your laptop.

## Start here

<CardGroup cols={2}>
  <Card title="Quick start" icon="rocket" href="/quickstart" horizontal>
    Install, connect, add an API key, and run your first AI task.
  </Card>

  <Card title="How Hopper works" icon="https://mintcdn.com/hypercubic/ssRtAUeGSvggjoxx/logo/hypercubic_logo_yellow.svg?fit=max&auto=format&n=ssRtAUeGSvggjoxx&q=85&s=99c2a91f0682082d0fea2957dd78e3cf" href="/core-concepts/how-hopper-works" horizontal width="391" height="391" data-path="logo/hypercubic_logo_yellow.svg">
    How the agent drives ISPF, JCL, and JES through one mainframe session.
  </Card>
</CardGroup>

## What it's for

Mainframes run core systems at most large banks, insurers, governments, and airlines. They use a 1970s block-mode terminal, JCL, and EBCDIC fixed-format files. Modern tooling expects HTTP, JSON, and UTF-8.

Common mainframe requests and how Hopper handles them:

| Request                              | Hopper's actions                                                    |
| ------------------------------------ | ------------------------------------------------------------------- |
| Read this dataset                    | FTP retrieve, EBCDIC-to-ASCII transcode, return UTF-8 text          |
| Submit this job                      | Generate JCL, submit via JES, poll, fetch spool                     |
| Press PF3                            | Send the TN3270 attention key, wait for host response               |
| Decode this VSAM record              | Parse the COBOL copybook, decode each field, return structured data |
| Run a SQL query over production data | Snapshot to a local file, query that                                |

## What the chat agent does

You ask in plain English. The agent picks the right operations and runs them.

* "Show me what `CBACT01C` does." It reads the COBOL from the source PDS and walks you through it.
* "This job failed last night — what happened?" It pulls the spool, classifies the abend, points at the offending step.
* "Add a check for negative balances to the validation paragraph." It reads, edits, recompiles, reports back. You approve before any write.
* "How many accounts have a balance over \$1M?" It snapshots the VSAM file, decodes via copybook, runs SQL.

Mutating actions like writing a dataset, submitting a job, or modifying CICS resources pause for your approval. Approve to run, reject to cancel. See [Tools](/core-concepts/tools).

## Who uses it

* Mainframe developers who want a modern interface and an AI assistant that understands JCL, COBOL, and CICS.
* Engineers onboarding to a mainframe codebase, who need to read and reason about old code without years of context.
* Modernization teams encapsulating mainframe systems behind modern APIs.

## Where it fits in modernization

Mainframe modernization usually means rehosting, rewriting, refactoring, or encapsulating. Hopper is for the encapsulating phase: read what's running, understand the data, test interactions before you commit to a direction.

It does not migrate, rewrite, or translate COBOL.

## What Hopper doesn't do

* **No mainframe-side install.** The product runs on your laptop.
* **No bulk data movement.** Hopper reads and writes datasets; it doesn't move data between systems.
* **No production change management.** The deploy feature targets dev and test environments.

## Learn the basics

<CardGroup cols={2}>
  <Card title="How Hopper works" icon="https://mintcdn.com/hypercubic/ssRtAUeGSvggjoxx/logo/hypercubic_logo_yellow.svg?fit=max&auto=format&n=ssRtAUeGSvggjoxx&q=85&s=99c2a91f0682082d0fea2957dd78e3cf" href="/core-concepts/how-hopper-works" width="391" height="391" data-path="logo/hypercubic_logo_yellow.svg">
    How the four panels share one mainframe session.
  </Card>

  <Card title="The AI agent" icon="robot" href="/core-concepts/the-ai-agent">
    Turns, tool calls, history, and approvals.
  </Card>

  <Card title="Tools" icon="shield-check" href="/core-concepts/tools">
    What the agent can do, and how Hopper keeps you in control.
  </Card>
</CardGroup>

## New to mainframes?

If you've never seen TN3270, JCL, or EBCDIC, start with [How Hopper works](/core-concepts/how-hopper-works). The chat agent can also explain the screen you're on. Paste or attach it and ask.