01 · Install

Install Lyriks with one command.

The installer checks your machine, proposes what is missing, installs Lyriks in containers and verifies the result.

Prerequisites

  • A machine x86-64. Apple Silicon Macs run it through Rosetta.
  • An operating system Linux, macOS, or Windows with WSL2.
  • Docker and Compose Installed for you if missing (Docker Desktop on Windows).
  • An LLM of your choice Hosted, local or sovereign, with your own tokens.
  • One outbound flow, at install only HTTPS to registry.lyriks.io to pull the images. In isolated environments, the images can be delivered another way.

The command

Linux & macOS

$ curl -fsSL https://get.lyriks.io | sh

Read the script first: get.lyriks.io/install.sh

Windows · PowerShell

> irm https://get.lyriks.io/windows | iex

Read the script first: get.lyriks.io/windows

What the installer does

  1. 01Checks the machineArchitecture, operating system, WSL2 on Windows, Docker and Compose.
  2. 02Proposes what is missingRequired packages are installed if needed.
  3. 03Installs LyriksEverything runs in containers, database included.
  4. 04Verifies itselfIt reports success only if the installation works.

Isolated or air-gapped network? Lyriks needs no outbound connection to run. Only the images come from registry.lyriks.io at install time, and they can be delivered another way.

Plan an isolated install with us

Details for every system: install guide. Something went wrong? Troubleshooting.

Get your licence key

02 · Use Lyriks

Your first 30 minutes.

Start small, on one module that matters. You will see what Lyriks recovers before you commit to anything bigger.

  1. 01Open LyriksGo to http://localhost:3000 and claim your installation with the email used for your key.
  2. 02Start with one modulePick a module that matters, not the whole system. Point Lyriks to its code, its docs, or both.Generate a spec from code →
  3. 03Read the 360° specFeatures, journeys, rules, screens and data, each rule linked to the line that implements it.
  4. 04Validate with your teamConfirm, correct or drop each rule. Inconsistencies between sources show up with their sources.
  5. 05Connect your coding agentSend the validated spec over MCP to Claude Code, Cursor, Copilot or any MCP agent, then read the implementation report.Use Lyriks with your agent →

The full walkthrough, with commands and a quick test: First steps.

Guides by use case

See it live

03 · Q&A

The questions teams ask.

The product

What does Lyriks do, in one sentence?

Lyriks reverse-specifies your existing code, lets you refactor or evolve it while measuring the impact of every change, then verifies the result against the spec with a deterministic engine.

How is it different from LLM-generated documentation?

An LLM produces text: documented, but neither verifiable nor linked to the code. Lyriks builds a structured spec, in a graph, where each rule points to its source and to the code that implements it. You can validate it, simulate it and check the code against it.

How is it different from AI code review?

Code review reads the code line by line once it is written. Lyriks verifies the spec before implementation, then checks the delivered code against it. The two approaches complement each other.

Why “deterministic”?

The Lyriks verification engine does not use AI: it computes. The same question always gets the same answer, which makes every verdict auditable.

Which languages and projects?

Lyriks reads your existing code and documentation, including older systems. The simplest way to know for your case: try it on one module, or talk to us in a demo.

LLMs and agents

Do we need an LLM?

Yes. Lyriks relies on the LLM of your choice to read code and produce code. Verification itself is done by the deterministic engine.

Which agents does Lyriks work with?

Claude Code, Cursor, Copilot and any MCP-compatible agent. Lyriks sends them the spec, its rules and their sources.

What if we switch models?

Project memory is independent of the LLM. Models come and go, memory stays.

Security

Does our code leave our premises?

No. Lyriks installs on your servers, can run air-gapped, and sends nothing beyond what you already give your LLM, without your permission.

Is Lyriks open source?

Yes, the engine is published under the AGPL.

Installation and pricing

How do we install Lyriks?

With one command on Linux, macOS or Windows. The script checks the machine, offers what is missing, installs Lyriks in containers and tests itself.

How much does Lyriks cost?

The Community Edition is free. Solo Builder is $55 excl. tax per project per month, Team $200 excl. tax per project per month. For Enterprise, contact sales. Online payment is coming soon: book a demo and we set up your plan with you.

Can we start small?

Yes, we recommend it: a pilot on one or two modules, to prove the value on a concrete case before scaling.

See the full FAQ, or more on security and pricing. Not sure where to start? Take the two-minute diagnostic.

04 · Studies

Take it with you, share it with your team.

Free documents to read offline and forward. See what each one covers, then leave your work email and the PDF downloads right away.

For CTO, engineering manager, tech lead

Guide · free · 11 pages

The legacy takeover checklist

What to inventory, recover and decide before you change code nobody understands anymore. Checklists you can tick, and a 30/60/90-day plan.

You inherited a system that runs, that customers depend on, and that nobody fully understands anymore. This is a working checklist: print it, tick the boxes, hand pages to your leads. The goal is concrete. In about ninety days, know one or two parts of the system precisely, agree with the business on what they should do, and ship a first change you can stand behind.

Inside

  • What to inventory before you touch anything: code, environments, written sources, people
  • How to recover the spec of a module and find where the sources disagree
  • How to secure the first change, and the decisions to take with the business
  • A 30/60/90-day plan, and seven common mistakes to avoid

For CTO, head of engineering, QA lead

White paper · free · 15 pages

Verifying AI-written code: a deterministic approach

The verification gap in numbers, the three gaps to look for, why the check must be deterministic, and how a verification loop fits next to code review.

AI assistants have sharply increased the volume of code that reaches review, and trust in that code is low. Code review can judge how code is written. It cannot tell whether the code does what the business decided, because that decision lives outside the code. This paper explains the missing reference, the spec, and how a deterministic verification loop gates a release on it.

Inside

  • The verification gap, with published industry data
  • The three gaps to look for: additions, omissions, regressions
  • Why a check that gates a release must be deterministic
  • How the verification loop works, where it fits in the pipeline, and how to run a pilot

For CISO, infrastructure team

Security pack · free · 14 pages

Lyriks security pack, for CISOs

Deployment model, network flows, data and the LLM question, plus pre-filled answers to the usual security questionnaire.

Written for the people who approve, install or audit a tool before a team can use it. It sticks to what Lyriks states publicly about its product and its installation. When a question is not covered, it says so and tells you whom to ask, instead of filling the gap with assumptions.

Inside

  • Six facts for the CISO, starting with: it runs on your servers, with your LLM
  • Deployment model, network flows, and the data Lyriks reads, produces and stores
  • The LLM question, the open source licence and auditability
  • Pre-filled answers for your security questionnaire, and a deployment review checklist

Keep your software in tune

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