Introduction

Why Lytos?

You use AI to code. Without a shared reference, every session tends to start from scratch — re-explaining the project, the constraints, the conventions. The AI forgets, you repeat yourself, the output can feel generic.

You may have already tried a persona for your agent — “You are a Senior Developer with 20 years of experience.” It’s a common approach that mostly shifts the tone. The effect on actual quality feels limited to us.

Lytos offers a complement: a frame the AI can read at every session.

“Without a shared frame, every session starts from zero. With Lytos, every session can build on the last.”

Rather than dressing up the AI, you give it what it most often lacks over time: context that persists, procedures that are precise, and quality criteria that are verifiable — as a complement to the practices you already have.

The 5 pillars

Intent

Your project’s constitution — the why, the stack, the principles. Read by the AI at every session.

Design

Reusable procedures — code review, testing, deployment. The AI follows steps, not vibes.

Standards

Quality criteria — files under 300 lines, no magic numbers, tests required. Verifiable, not subjective.

Progress

Issues and sprint — what’s moving, what’s blocked. The kanban board drives the work.

Memory

Persistent brain — architecture decisions, patterns, bugs solved. The AI remembers across sessions.

The idea

A large part of the AI industry copies human organization — agents with personas, roles, titles. Lytos proposes a complementary shift: rather than managing agents, you can define the framework in which they operate.

The image we keep is that of Kubernetes pods — stateless, scalable, without their own identity. You don’t name a pod: you define the desired state and the orchestrator allocates resources.

Works with everything

Any language: Python, JavaScript, TypeScript, Go, Rust, PHP, Swift
Any AI tool: Claude Code, Cursor, OpenAI, Codex, and whatever comes next
Any project size: from solo side project to 50-person team