Afunana reads your COBOL, RPG, CL, and JCL source code and turns it into living, interactive documentation — business narratives, flowcharts, dependency maps, data dictionaries, and impact analysis. Ask questions in plain language. Get answers with source-line citations. Describe a change. Review the plan. Execute it on IBM i or z/OS. No IDE, no consulting engagement, no black box.
Core business systems run on IBM i and z/OS. The developers who built them are retiring. The business logic lives only in the code, understood by fewer people every year. Every change is a risk taken blind.
A production bug takes days to diagnose when nobody understands the business logic hiding in COBOL. Downtime costs money and erodes trust.
Every change to legacy code is a leap of faith. You can't see what will break. You can't trace what depends on what. Every deployment is a gamble.
Every retirement removes institutional knowledge that exists nowhere else — not in docs, not in wikis, only in the code and in one person's memory.
Afunana closes the loop between understanding your code and changing it. Every stage feeds the next. Documentation stays current because every change automatically triggers re-analysis of the programs it touched.
Business rules, call trees, field references, six types of flowcharts, data dictionaries, dead code flags — all extracted from source, all traced to the line.
Plain-language intent becomes an exact plan: which members, which lines, which compile order, which regressions to run. Every edit cited to the source line it will modify.
One click writes the change directly to the machine. Live log streams every step as it runs. Nothing happens without your explicit approval.
Programs touched by the change get re-analyzed. Delta rebuild — only affected programs, not the whole collection. Your documentation never goes stale.
Afunana reads your source and produces documentation no human has the time to write. Not a transcript of the code — the business reason it exists, what breaks when it changes, and why.Accessible to developers, analysts, managers, and auditors — no IDE required.
A business-level summary of what the entire system does, its major functions grouped by purpose, and how everything connects — generated from the actual source code.

Searchable, filterable lists of every program (COBOL, RPG, CL, JCL) and every file. Each entry carries a business-level description, auto-generated tags, and type indicators — not a signature dump.

Every program gets a full documentation package: business summary, technical details, dependencies, and six interactive diagrams. Each finding anchored to the source line that justifies it.

Every field with its technical type and length — plus the business meaning, the consequences of corruption, and the programs that rely on it. Physical and logical file relationships mapped. Key fields identified.

Every program, every call, every file dependency — in one interactive tree. Click any node to see its documentation inline. Expand, collapse, search, and navigate a system of thousands of objects without opening a single file. Built from the full IBM i / z/OS system catalog — programs, files, fields, call chains, key structures, and procedure cross-references all linked and traversable.

This is what makes Afunana more than a documentation tool. Describe the change in English or Hebrew. Review the plan. Approve it. Watch it run on your machine. Nothing executes without your explicit approval.
Ask in plain language — "change the national ID from 9 to 10 digits" — and Afunana proposes the full change: which members to edit, which lines, which compiles, which regressions to run. Nothing happens without your approval.

Plans are persisted with an ID, a summary, a step list, and a status. Pending. Approved. Failed. Completed. You can open any plan, inspect every step, and decide whether to run it — today, tomorrow, or never.

Every step shows its type (source edit, compile, file change), the object it touches, the exact program, the source member, the line number, and what will change. Compile order and rollback capability computed automatically. IBM i CL commands or z/OS JCL included — not generic build instructions.

Afunana isn't just for code you already have. Start a new program from a template, from an existing program, from a file structure, or by uploading an analysis document. Every step is editable — AI-filled or hand-written, you choose.
Source → Files & fields → Lookup tables → Validation → Screens → Processing → Messages → Review & generate. Output is compilable COBOL or RPG, ready to deploy with the same Plan & Execute pipeline.

Four tool families, all working against the same intelligence layer. No re-analysis, no re-upload, no separate products — everything connects.
Generate new programs from templates, existing programs, file structures, or analysis documents. Edit live source on IBM i or z/OS with documentation alongside. AI-assisted modifications with diff view.
Change impact analysis — see what breaks before you touch it. Dead code detection with risk assessment. Cross-reference reports for any field, file, or program. Embedded SQL and QSQLSRC table access mapping. System health dashboards.
Business-level documentation for every program and file. Data dictionaries. System overviews. Six types of interactive flowcharts. Auto-generated business taxonomy and tagging. Auto-refreshing documentation that stays in sync with every change.
Built-in Categories → Epics → Tasks workflow with Kanban board. Comments and assignments. Analysis findings flow directly into actionable work items — no external tool needed.

Afunana ships with everything needed to operate a real system: live admin panel, full audit trail, configurable AI, build pipeline management, user and role management, bilingual UI. A non-engineer can operate the platform.
One admin panel for collections, users, projects, audit logs, change plans, prompts, configuration, and deployment. No SSH, no config files, no restarts.

Create, configure, and build each collection. Per-collection output language, user access, and build history.
Admin, User, Viewer, QA. Role-based access to collections. Password management from the UI.
Built-in Categories → Epics → Tasks workflow with Kanban board for managing work between teams.
Every admin action — build, deploy, plan, config change — logged with user, action, target, timestamp, and IP.
Every AI prompt editable live with placeholder validation. Changes take effect on the next call — no restart.
Every backend setting editable, grouped by category, with secrets masked.
Trigger a build from the admin; watch it run. Five-stage pipeline: parse & chunk, LLM documentation, system artifacts, embed & index, auto-tag. Every stage timed, every result captured. Delta rebuilds regenerate only affected programs.

Change the LLM behind any pipeline stage from the admin — no code changes. Primary + fallback chains across providers. Temperature-zero classification for deterministic, reproducible results. Multi-layer caching: query shortcuts, response cache, prompt cache. Hard token budgets on every role. Build once, query many — the expensive analysis happens once; every subsequent query is near-zero cost.

Version bump, build, push, trigger — all from the admin, all streamed live. Deploy history kept with git commit tracking. Rollback is straightforward.

Most legacy tools live inside an IDE — locking out everyone who doesn't write code. Afunana runs in a browser. Developers, analysts, managers, and auditors all see the same system, the same documentation, the same evidence.
Understand your full legacy estate. See dependencies before committing to changes. Make modernization decisions on data — not guesswork. Prove compliance with full documentation and audit trails.
Business rules, process flows, and system overviews in plain language. No IDE needed. Open a browser and navigate the system the same way developers do — without touching source code.
See exactly what breaks before a change goes in. Approve plans before they run. Assign work using built-in project management. Onboard new developers in days — not months.
Business rules, dependencies, and impact analysis in plain language. Six types of flowcharts. Describe the change — get an exact plan back. Execute with one click. Optional VS Code extension for in-IDE documentation.
Every finding traced to a source line. Complete system documentation. Full audit trail of every admin action. Cross-reference reports. Change history with before/after evidence.
Every finding traced to a source line. Every change plan shows exactly what will happen — member, line, compile command — before you approve it. AI that cites its sources, operates deterministically, and asks before it acts.
Every plan is reviewed and approved before it executes.
Every factual claim includes a source program and line number.
Classification and routing are deterministic. Same question, same path, every time.
Deploy on your own infrastructure. Source code never leaves your network. The Afunana team has no access to your server, your data, or your LLM. No telemetry, no phone-home, no remote management. You choose the LLM — on-premise models, your Azure tenant, your AWS account, or any API.
No VS Code setup. No extension management. No per-developer configuration. Afunana runs in any browser. Developers, analysts, managers, and auditors all use the same platform. For developers who prefer their IDE, an optional VS Code extension brings documentation into the editor — as a complement, not a requirement.
Each team works on their own collections with role-based access. Hebrew and English UI with full RTL support. Horizontal scaling — dedicated instances per collection for large deployments. Admins manage builds, models, users, and execution approvals from one panel.
Afunana turns legacy from a burden into an asset — and from something you read into something you can change.
IBM i · z/OS · AS/400 · COBOL · RPG · CL · JCL · SQL · One platform. No IDE required.
Schedule a demoTell us about your environment — how many programs, which languages, what you're trying to understand or change. We'll show you working results on your own code in days, not months.