Reverse-specify. Evolve. Repeat.

Understand any software.
Evolve it at AI speed, at no risk.

Lyriks recovers the spec of your code, simulates every change, and checks the result rule by rule. Deterministic: same code, same verdict, every run.

JOURNEYSFEATURESRULESSCREENSDATAARCHITECTURE

The problem

AI writes twice as much code. Only half of developers check it.

+98%more pull requests merged in teams coding with AIFaros AI, 2025
48%of developers always check AI code before approving itSonar, 2026
29%trust the accuracy of AI, down from 40%Stack Overflow, 2025

The result: nobody dares touch the existing code anymore.

Legacy code has lost its spec, generated code goes unreviewed, and every change is made blind.

Use cases in practice

Five ways to keep your software in tune.

01 · Take over legacy code

The code nobody dares touch, finally explained.

Lyriks reverse-specifies your software: every rule, screen and journey recovered and linked to the code that runs it.

See the use case

02 · Refactor and migrate

Rebuild without losing a single rule.

Old language, aging database or monolith: your LLM rewrites from the validated spec, and Lyriks checks that every rule made the trip.

See the use case

03 · Evolve safely

Every change simulated before it is coded.

Your team runs several changes at once. Lyriks shows what fits and flags conflicts before the first line.

See the use case

04 · Verify AI code

Provable QA, rule by rule.

After each run, a deterministic report: what conforms, what was added, left out or broken. Gaps go back to the agent.

See the use case

05 · Project memory

Models come and go. Memory stays.

Spec, rules and decisions traced to their source, on your servers. Switch LLM whenever you like.

See the use case
Not sure where to start?Take the two-minute diagnosticEight questions, a control score, and the use case to start with.Start the diagnostic

Security

Nothing leaves your premises.

Security

On your servers

On-premise, air-gapped, with no outbound connection.

Your LLM, your tokens

Lyriks sends nothing beyond what you already give your LLM. Local or sovereign models work.

Memory you own

Exportable in full, with its traceability. No lock-in.

An open engine

Open source under the AGPL: inspect it, install it with one command.

In numbers

8,000+downloads since May
14,000+projects
1.7M+features recovered
3M+inconsistencies and conflicts flagged

Estimates based on active installations and typical project size.

About Lyriks

Software engineers. Shipping product for 15 years.

Lyriks is built by software engineers who have been shipping product for 15 years. Its engine, built in-house on graph rewriting, computes rather than guesses whether code does what was decided.

Around that engine, engineers, PhDs in science and AI, and people who have run consulting divisions and AI programs. We build Lyriks for the teams who no longer dare touch their own code.

  • Adrien Basso-BlandinCEO
  • Adrien CharlesCPTO
  • David MabbouxCRO
  • Louis ChevrelCustomer Success Manager
Meet the team

Pricing

Free to start. Simple to grow.

Monthly prices, excluding tax. Online payment is coming soon: book a demo to get started.

See pricing

Keep your software in tune

Start with one module. Lyriks recovers its spec, then verifies every change. All it needs is your repository.