Framework by David Offutt — 2026
The MICA Scale
Score and compare AI ecosystems across the four factors that determine real-world capability. Built for practitioners, not benchmarks.
M Model Intelligence
Raw LLM capability. Reasoning, accuracy, world knowledge, and instruction-following at scale.
I Integration Depth
MCP servers, plugins, Skills, and connectors. How deeply the AI touches your actual stack.
C Canonical Ability
Context window plus retrieval capacity. How much the system can read, hold, and reason across at once. Coined by David Offutt.
A Agentic Harness
The execution layer around the LLM. Slash commands, hooks, parallel sessions, file editing, terminal access, and error recovery. The gap between a chatbot and an agent.
Platforms
MICA Radar Chart
MICA Score Ranking