Kratos is a serverless intelligence platform for running ephemeral AI agents in isolated microVMs.
🎯 Core Vision: • Serverless compute platform for ephemeral agents • Micro-task execution: search, parsing, editing, content generation • Per-second billing model with strict resource limits • Energy-efficient compute optimized for ARM/Apple Silicon • Massively parallel execution without long-running processes
🏗️ Architecture: • Docker containers (temporary, will migrate to Apple Containers) • Lightweight, fast-launching environments similar to Firecracker • Ephemeral agents that spawn on demand and shut down immediately • Model-specific pulling for efficiency
🚀 Core API Functions:
- submit() - Deploy serverless agent for micro-task execution
- invoke() - Execute micro-tasks with burst efficiency
🔧 Technical Features: • Model extraction ensures only required models are pulled • Resource reclamation for efficient billing • Burst execution optimized for short-lived tasks
The system now embodies Kratos as "intelligence as a utility" - making AI accessible through efficient, ephemeral compute that scales like serverless functions but optimized for local, energy-efficient execution.