Service · Legacy modernisation
Legacy modernisation with AI agents
Coding agents can take on much of the reading, documenting and rewriting that a migration involves. What takes care is showing that the new system behaves like the old one, so we put the tests and review gates in place before the agents start.
Comonad modernises legacy codebases with AI coding agents in stages: characterise the current behaviour, capture it in tests, migrate behind those tests, and release through the gates of an AI-enabled SDLC.
The method is the one in our AI-enabled SDLC consultancy: plan first, route each task to the agent suited to it, and review every change against a gate before it merges. In a migration, the gate is behavioural equivalence with the system you already trust.
Your engineers stay in charge throughout. They own the target architecture and approve each stage, and they keep the tests and the method when we leave.
Tell us what the system is, what it is written in, and what you want it to become.
- CTOs and CIOs
- Heads of engineering
- Architecture leads
On this page
What agents change in a migration, and what they do not
Much of a migration is reading code nobody fully remembers, writing down what it does, and rewriting it piece by piece. Coding agents handle that volume of work well. What they cannot reliably judge is whether an odd behaviour is a bug or something a downstream system depends on.
So the order matters. First we capture what the system does today, in tests that run against the old code. Then agents migrate in small slices, and each slice has to pass the same tests before a person reviews and merges it. When we are unsure, the old behaviour wins until your team decides otherwise.
How work is planned and routed across agents is in Plan first, then route.
What an engagement covers
Assessment
The codebase, its dependencies, its interfaces and how it is tested today, and a recommendation on whether to migrate, wrap or leave each part.
Characterisation
Documentation and characterisation tests generated with agents and checked by your engineers, so current behaviour is written down before anything changes.
Staged migration
Agents migrate in small slices behind the tests, with review gates and a record of every AI-assisted change. Old and new can run side by side where the risk calls for it.
Handover
Your team keeps the tests, the gates and the method for the rest of the estate.
What you keep
Assessment
What to migrate, wrap or leave, and in what order.
Characterisation tests
Current behaviour, captured before any change.
Migration record
Every AI-assisted change, its review and its tests.
Method
Routing, gates and runbooks your teams can reuse.
How an engagement runs
Assess
With the engineers who maintain the system.
Characterise
Tests that pin today's behaviour.
Migrate
Small slices, each through the gates.
Hand over
Your team carries on with the same method.
What we do not do
- We do not give legal advice. Legal sign-off stays with your counsel.
- We do not resell a model, an agent platform or a coding assistant, and we take no referral fees.
- We do not migrate any part of a system until its current behaviour is captured in characterisation tests.
- We do not merge AI-written changes without review by your engineers.
Who it is for
- CTOs and CIOs carrying legacy systems that are expensive to change or hard to staff.
- Engineering teams that want the speed of agents without losing control of behaviour.
- Organisations already setting up an [AI-enabled SDLC](/services/ai-sdlc) that want a first large piece of work for it.

Alex Vakhitov
Founder, Comonad · AI and software architect, London
Alex Vakhitov founded Comonad in 2024. An AI and software architect, he was co-founder and CTO of Cloud KB, UK energy-market software, from August 2015 to March 2022.
Questions about legacy modernisation
Which languages and platforms?
The method does not depend on the language. What matters is whether the old system can be run in a test environment, so its behaviour can be captured. We confirm that in the assessment.
Which coding agents do you use?
Whichever your policies allow. We usually route tasks across more than one, and we are not tied to any vendor.
How long does it take?
It depends on the size of the system and how much of it is tested today. The assessment gives you a staged plan before you commit to the migration.
Related
Discuss a modernisation
Tell us what the system is, what it is written in, and what you want it to become.
Discuss a modernisation