Definition

What is Agent Drift?

Last updated

The gradual behavioral deviation of an AI agent from its original goal, role, or correct operation over long-running tasks.

Agent drift is the umbrella phenomenon covering six distinct failure modes in long-running agentic systems: goal drift, context drift, role drift, tool-use drift, hallucination cascades, and plan decay. Most production agent failures trace to one of these mechanisms rather than to raw model capability, and most are solvable with context engineering rather than bigger models.

Related Context Drift Context Rot AI Agent Multi-Agent System Context Engineering

Further reading

Articles about Agent Drift

Agent drift: why long-running AI agents lose the plot

Agent drift is how AI agents silently deviate from goals over long-running tasks. Six mechanisms cause it, and most have nothing to do with the model.

All terms

View full glossary
Agent Drift AI Agent AI Hallucination Context as a Service Context Container Context Compression Context Drift AI Second Brain Context Engineering Context Poisoning Context Portability Context Rot Context Window Epistemic Provenance MCP Server Fine-Tuning MCP (Model Context Protocol) Prompt Caching Prompt Engineering Multi-Agent System RAG (Retrieval-Augmented Generation) Semantic Search Structured Context Wire

Put context into practice

Create your first context container and connect it to your AI tools in minutes.

Create Your First Container