What Does This Mean?
The tooling layer built around a base LLM to make it act as an agent — its tool definitions, prompt structure, memory handling, and control loop — assembled before the agent ever runs, as distinct from the model itself or the runtime that executes it.
REAL-WORLD EXAMPLE
"The model didn't change at all — we just rewrote the scaffolding and it stopped hallucinating tool calls."
LORE & ORIGIN
Standard 2026 AI-engineering vocabulary distinguishing what a model is (weights, training) from what it's built with to act agentically. Documented in practitioner glossaries (Hugging Face, zbrain.ai) and formalized in academic work like the 2026 arXiv paper "Inside the Scaffold: A Source-Code Taxonomy of Coding Agent Architectures." Increasingly discussed alongside the related but distinct terms "agent harness" (the runtime orchestration layer) and "agent runtime" (the infrastructure the harness runs on).
MENTIONS OVER TIME
Radar observation data not yet available for this term.
RELATED SEARCHES
TOP PLATFORMS
Not enough Radar data yet to show a platform breakdown.
WHEN DID YOU FIRST HEAR THIS?
CLICK AN OPTION BELOW TO CAST YOUR VOTE.
[ VOTE TO REVEAL COMMUNITY RESULTS ]