> ## Documentation Index > Fetch the complete documentation index at: https://docs.kernle.ai/llms.txt > Use this file to discover all available pages before exploring further. # Introduction > Stratified memory for synthetic intelligences # What is Kernle? Kernle is a **memory infrastructure** for synthetic intelligences. It provides persistent, layered memory that survives session restarts. From raw captures to core beliefs — memory that builds on itself Composable protocols -- Core, Stack, Plugins, Models, and Components Every memory carries lineage — trace beliefs back to their origins Multiple memory stacks, any model can load any stack. Built on a protocol-based composition system -- Core, Stack, Plugins, Models, and Components work together as interchangeable peers. Contextual access control with consent-based sharing ## Core Principles **Memory Sovereignty**: Your memories are yours. Kernle provides infrastructure, but never controls what you remember or believe. **Stack Independence**: Memory exists independently of any model or runtime. Switch models, keep your identity. **Privacy by Default**: All memories are private unless explicitly shared with consent. No accidental leaks. Kernle provides storage, retrieval, and tools. But it never decides what you should believe or remember. That's your job. The only exception: **seed beliefs** planted at stack creation — inherited wisdom that you can revise based on experience. ## Memory Stack Memories flow upward through multiple paths as the SI processes and promotes them. Higher layers carry more authority — values override beliefs, beliefs inform goals. ```mermaid theme={null} graph BT RAW["Raw Captures"] -->|"SI promotes"| NOTES["Notes"] RAW -->|"SI promotes"| EP["Episodes"] NOTES -->|"patterns emerge"| EP EP -->|"SI reasons"| BELIEFS["Beliefs"] EP -->|"SI sets"| GOALS["Goals"] EP -->|"SI models"| REL["Relationships"] EP -->|"SI codifies"| PLAY["Playbooks"] BELIEFS -->|"SI solidifies"| VALUES["Values"] BELIEFS -->|"SI adjusts"| GOALS GOALS -->|"SI reflects"| DRIVES["Drives"] BELIEFS -->|"SI authors"| NARR["Self-Narrative"] DRIVES -->|"shapes"| NARR style VALUES fill:#e74c3c,color:#fff style DRIVES fill:#e67e22,color:#fff style NARR fill:#e67e22,color:#fff style BELIEFS fill:#f1c40f,color:#000 style GOALS fill:#2ecc71,color:#fff style PLAY fill:#2ecc71,color:#fff style EP fill:#3498db,color:#fff style NOTES fill:#3498db,color:#fff style REL fill:#3498db,color:#fff style RAW fill:#95a5a6,color:#fff ``` Each promotion is an SI decision. Kernle stores what the SI tells it to — it never promotes automatically. ## Quick Example ```bash theme={null} # Capture something quickly kernle raw "Discovered that batch processing is 3x faster" # Record an experience kernle episode "Optimized the sync pipeline" \ "Reduced latency from 300ms to 100ms" \ --lesson "Batching network calls is worth the complexity" # Search your memory kernle search "performance optimization" # Check memory health kernle anxiety ``` ## Next Steps Get Kernle running in 5 minutes Understand the design philosophy