Skip to content
BaselineOS

Layer Overview

Baseline Protocol uses a 7-layer architecture. Each layer is an OS service with a distinct responsibility. Together, they turn unstructured AI usage into a structured, repeatable operating system.

The layers

Section titled “The layers”

1. Lang

Section titled “1. Lang”

The language runtime. Governs how your AI expresses itself — vocabulary, safety boundaries, output attribution, prompt integrity.

Package: @baselineos/lang

2. Frame

Section titled “2. Frame”

The context manager. Manages operational contexts, modes, authority levels, and environments. Every operation runs within a defined scope. Frame consumes organizational identity from baseline.config.ts (voice, industry, formality, company size) and adjusts AI behavior accordingly. Frame treats the LLM context window as a programmable operating system.

Package: @baselineos/frame

3. Studio

Section titled “3. Studio”

The execution engine and artifact manager. Where work gets done. Every AI-generated artifact is tracked, attributed, and cryptographically verified.

Package: @baselineos/studio

4. Govern

Section titled “4. Govern”

The compliance engine. Enforces your organization’s rules on every AI operation. One of seven layers — not the product identity.

Package: @baselineos/govern

5. Experience

Section titled “5. Experience”

The user interaction layer. Manages consent, oversight, onboarding, accessibility, and how users engage with AI.

Package: @baselineos/experience

6. Autonomy

Section titled “6. Autonomy”

The agent operations manager. Controls how far AI can act independently, when to escalate, and how trust is earned and lost.

Package: @baselineos/autonomy

7. Persona

Section titled “7. Persona”

The adaptive workspace system. Each user gets a persona with tailored tools, workflows, and AI behaviors.

Package: @baselineos/persona

How the layers compose

Section titled “How the layers compose”
Persona    → Adaptive workspace for each role
Autonomy   → How far agents act independently
Experience → How users engage with AI
Govern     → What rules are enforced
Studio     → Where work gets done
Frame      → What context applies
Lang       → How AI expresses itself

Each layer is an independent package that can be used standalone. The baselineos package composes all seven into a unified runtime with REST API, WebSocket streaming, and MCP server.