AI-assisted deterministic Java refactoring at scale

A.I. agents are now part of your development team. However they don’t grasp the intricate details of your existing codebase or guarantee safe code transformations at scale.

CodeLaser combines both and gives your A.I. refactoring superpowers.

Built on a functional model of your code

CodeLaser doesn’t work on syntax. It builds a full functional model of your Java codebase that all transformations can use. Code transformations are expressed by intent, not by syntax changes. The result is maximum refactoring power with minimal coding. Think a few lines of code that split a 200-method class into cohesive parts while preserving dependencies, fields and parameters lists. Instantly, without errors.

Static analysis tools find problems but don’t fix them.
Syntax-based transformers handle known recipes but don’t understand structure. A.I. agents write code but can’t guarantee safety across a dependency graph. CodeLaser bridges the gap: structural analysis and deterministic refactoring, working with your AI, not instead of it.

Let your A.I. agent do the work

CodeLaser seamlessly connects to any A.I. agent. Give your agent superpowers to analyse and refactor at scale by writing and running CodeLaser flows. Discuss flow results during agent conversation as you refine scripts and flows that can eventually be deployed in your CI/CD pipeline for continuous improvement.

Visual pipelines, not black boxes.

Build multi-step refactoring workflows as directed graphs. Each node is a discrete step — configure, run, inspect results inline. Deploy detection and remediation flows in your CI/CD pipeline to keep your code clean and modern going forward.

Script, Repeat, Deploy.

Write analysis and refactoring logic in Starlark (Python) against the complete CodeLaser code model — types, methods, call graphs, dependencies. Version-controlled, reproducible, team-shareable. Use MCP to let your A.I. agent write code for what you want to achieve.

Map every dependency. Find every cycle. Then fix it.

Turn dependencies into actionable data. See where your architecture is coupled and has cycles — and trace exactly why. Use CodeLaser’s dependency optimization and code relocation primitives to optimize your dependency structure.

See what’s dead. Remove it safely.

Reachability analysis from real entry points. Dead code highlighted in source, clustered into dependency trees. Remove entire groups in one operation, down to the minute rippling effects in parameters, types and interfaces. All with a few powerful statements, and extensive human review where needed.

Get a demo

A.I. assists understanding, planning, and decision-making. All code transformations are executed by the deterministic engine. Same input, same correct output — every time.

We founded CodeLaser based on our own experiences with rapidly changing Java codebases, and the lack of tools to support surgical refactoring at scale. We are developers building for developers.
Bart Naudts
Founder, Head of Technology