Self-healing AI conversion pipelines

Modernize legacy code, SQL & reports
without the rewrite risk.

Code Alchemist converts legacy code, database SQL and report definitions between technologies using AI agent pipelines that verify their own work and heal themselves when a step falls short.

Request a demo See how it works

Verified conversions (UVM)
Multi-tenant & secure
Bring your own LLM keys
Air-gapped deployment

The problem

Legacy modernization is slow, risky and expensive.

Rare skills

COBOL, RPG, VB6 and hand-tuned SQL live in the heads of a shrinking pool of experts.

Manual & error-prone

Hand-porting dialects and report formats is tedious, and subtle behavior changes slip through.

Hard to trust

“Did the conversion actually preserve behavior?” is the question that stalls every project.

What it does

One platform, many conversions.

Purpose-built system pipelines for the migrations enterprises actually face, each one a chain of specialized AI agents tuned for the source and target technology.

SQL Conversion

Oracle · SQL Server · DB2 PostgreSQL · DB2 · MySQL

Migrate stored procedures and queries across database dialects, then verify the converted SQL returns the same row counts against a live connection.

BIRT SQL Conversion

BIRT .rptdesign Target dialect

Re-target the SQL embedded in BIRT report designs to a new database dialect while preserving report structure.

Java to Auto Scripts

Java automation Maximo Jython

Convert Java automation scripts into IBM Maximo automation (Jython) scripts, ready to drop into the platform.

BIRT ↔ Jasper

BIRT .rptdesign JasperReports .jrxml

Bidirectional report-format modernization with element-parity checking so nothing is dropped in translation.

COBOL → Java

COBOL programs Java

Translate legacy COBOL business logic into maintainable Java, with analysis and complexity estimation up front.

Stack Modernization

RPG / RPGLE · VB6 Java · C# / .NET

Re-platform an entire legacy stack to a modern one, moving RPG or VB6 to Java or C#/.NET, with a configurable source-to-target pipeline covering backend, frontend and report engine.

QRadar → Sigma

QRadar SIEM content Sigma / Roota rules

Convert vendor-specific SIEM detection content into vendor-neutral Sigma detection rules.

How it works

Every conversion runs as a verified agent pipeline.

Upload a source file and it flows through a graph of AI agents. Each stage does one job well, and the final stage checks the result against the original.

01

Analysis

An agent reads the source file and documents its structure, dependencies and intent.
02

Complexity

A complexity-analysis agent scores the effort and risk, informing model selection.
03

Conversion

The conversion agent produces the translated code, SQL or report, keeping it dialect-aware and structure-preserving.
04

UVM Verification

The Universal Verification Module inventories source vs. output elements, runs behavioral checks and scores fidelity.

The differentiator

Pipelines that grade and heal themselves.

Most tools convert and hope. Code Alchemist holds every agent to a measurable quality contract, and when one falls short it fixes that step precisely, then verifies again.

GREEN AMBER RED

Quality contracts

Every agent is graded against declarative acceptance criteria: valid syntax, element parity, and behavioral checks like matching row counts against a real database.

Targeted healing

When a node underperforms, the engine re-runs just that node with feedback (retry, escalate to a stronger model, chunk the input, or route to a human) instead of blindly re-running the whole job.

Confidence verdicts

Each run emits a scorecard with a GREEN / AMBER / RED confidence verdict and a coverage percentage, so you know exactly how much to trust each output.

Healing memory

Proven fixes are recorded by problem signature and re-applied first when the same failure recurs, so the platform gets more reliable the more you run it.

Enterprise platform

Built for teams, tenants and trust.

Multi-tenant by design

An isolated Postgres database per tenant, with org-scoped LLM keys and role-based access (Admin / Analyst / Operator / Viewer).

Per-project DB connections

Connect source and target databases across PostgreSQL, MySQL, SQL Server, Oracle, MariaDB and IBM DB2, with named connection profiles (Dev / UAT / Prod).

Adaptive model selection

Assign the model each agent uses, or let complexity tiers auto-pick a model by program complexity to balance cost and quality.

Cost tracking & analytics

Per-call token and cost tracking with dashboards, plus a budget circuit-breaker so runs never blow past a spend threshold.

Human-in-the-loop sign-off

Configurable review checkpoints and a mandatory final approval keep a human in control of what ships.

Security-first

Encrypted credential storage, credentials never entering agent state, and SQL-injection / XSS screening on inputs.

Product tour

What you’ll actually use.

From your first project to your first verified output, and the workspace is organized around the way modernization teams work.

Projects

All your projects in one place

A searchable list of every conversion project, showing jobs, members, status and last activity, with one click to open one or start a new one.

Project

Everything for one conversion effort

Each project bundles Files, Pipelines, Jobs and Outputs behind a simple Create → Configure → Upload → Run → View workflow.

Pipelines

Configure a system pipeline

Pick a pipeline and see its agents (Analysis, Complexity, Conversion), plus optional UVM testing agents that execute and verify the output.

Jobs

Track every conversion run

Live job status with progress bars and per-job actions: resume a failed run from where it stopped, or cancel one in flight.

Job detail

Look inside a run

Follow each agent as it works, stream the logs, watch token cost accrue, and download the converted output the moment it lands.

Outputs

Collect every converted file

Download converted code, reports and generated documentation per job, grouped by type (coverage, analysis, converted scripts), or export them all at once.

Database

Connect source & target databases

Configure source and target databases per project across Oracle, SQL Server, DB2, PostgreSQL and more. Test credentials live and save Dev / UAT / Prod profiles.

API Keys

Bring your own LLM keys

Add per-provider API keys (OpenAI, Anthropic, Gemini…), isolated per tenant and never shown again after saving.

Healing rules

Self-healing correction rules

Auto-learned and expert-authored fixes, keyed by dialect and error code. Only proven rules are applied to new conversions.

Usage & Cost

Watch spend in real time

Organization-wide token usage and cost by model and project, a forecasted run-rate, and budget alerts so runs never blow past a threshold.

Screens shown are representative. Book a demo to see the live product with your own files.

Pricing

Plans that scale with your modernization.

Simple, credit-based pricing. Each plan includes a pool of conversion credits — run conversions until you use them up, then top up. No free tier, no metering surprises.

Loading plans…

Start in minutes — no sales call required. Prefer to talk first? Request a demo and our team will get you set up.

See it in action

Request a demo.

Code Alchemist onboards organizations by invitation. Tell us about your modernization workload and our team will set up a tailored walkthrough, and provision a tenant when you’re ready.

Demo video coming soon

Request a demo below for a live, guided walkthrough.

Getting started

Three steps to your first conversion.

1
Request access

Tell us about your modernization workload and the technologies involved.
2
We provision your tenant

A superadmin sets up your isolated organization and invites your team.
3
Connect & convert

Add your databases, upload source files, and run your first self-healing pipeline.

Request a demo Email us

Modernize legacy code, SQL & reportswithout the rewrite risk.