- Your developers keep the terminal and IDE they use today and point an open coding agent at a governed endpoint that runs inside your boundary.
- The agent searches your repositories, data dictionaries, interface specifications and design records, and cites what it used when it explains a module.
- Conversion runs in slices: explain the business rules, convert one module, generate tests that compare old and new outputs, then a developer reviews every change.
- H2O Cascade can fine-tune open code models on your own repositories, so the model learns your conventions without the code going to an outside service.
The code runs the mission, and few people can still read it
Your core systems still run on code written decades ago. The people who wrote it have retired or moved on, the documentation is thin, and every change carries risk because nobody is sure what else depends on it.
Your developers want AI help to read and convert this code, but your security team will not let source code, interface specifications or system designs go to an outside coding service, and in a classified or air-gapped enclave that service is not reachable at all. So the modernization backlog grows.
How legacy code conversion works
Step 1: Connect the coding agent to your endpoint
Developers point an open-source coding agent at a governed endpoint, which handles sign-in through your identity provider, role-based access and audit. Agent tool calls against source trees run in isolated execution.
Step 2: Index what the code depends on
Repositories, copybooks, data dictionaries, interface specifications, design records and runbooks go into private collections. Search finds exact program names, field names and error codes.
Step 3: Explain before you convert
The agent reads a module and writes the business rules it finds in plain language, with references to the source lines. A developer or analyst confirms the rules before any conversion starts.
Step 4: Convert one slice at a time
The agent converts a module into the target language and writes tests that run the old and new code on the same inputs and compare the outputs.
Step 5: Review and merge
Developers review every diff in their normal code review; nothing merges without a person, and the gateway keeps an audit log of the agent's requests.
Tuning on your own code
H2O Cascade adapts open code models to your repositories and conventions, so the model learns from your codebase without sending it to an outside service.