Reference
AI Context Glossary
Plain-English definitions for the key concepts in AI context management, MCP, and the Wire platform.
-
Agent Drift
The gradual behavioral deviation of an AI agent from its original goal, role, or correct operation over long-running tasks.
-
Agent Reliability
The consistency of an AI agent's output and success across repeated runs of the same task, rather than its performance on a single run.
-
AI Agent
An autonomous software program that uses a large language model to plan and execute multi-step tasks, typically by calling tools, reading data, and iterating on its own output.
-
AI Hallucination
A confidently-stated but factually incorrect output produced by a language model when it lacks grounded context for a claim.
-
AI Second Brain
A personal or team knowledge base built from notes, documents, and references that AI agents can query directly.
-
Chunking
Chunking is the process of splitting documents into smaller passages so they can be embedded, retrieved, and delivered to an AI model's context window.
-
Context as a Service
A model for delivering structured, AI-optimized context to agents and tools on demand.
-
Context Bloat
The accumulation of low-signal tool-call output and stale history that fills an AI agent's context window with content it no longer needs.
-
Context Compression
The practice of reducing token count in an AI agent's context window while preserving the information needed to complete tasks.
-
Context Container
A portable, shareable unit of organized context (documents, data, and structured information) made accessible to AI agents through MCP tools.
-
Context Drift
The gradual loss of task-relevant information as an AI agent's context window fills with accumulated history.
-
Context Engineering
The practice of deliberately designing, structuring, and managing the information provided to AI models to improve output quality and relevance.
-
Context Offloading
The practice of keeping an AI agent's working context window small by moving state to an external destination, such as a file, a sub-agent, or a retrievable store, and bringing it back only when a step needs it.
-
Context Poisoning
An attack in which malicious or false information is planted into an AI agent's memory, RAG index, or tool outputs so the model treats it as ground truth.
-
Context Portability
The ability to use the same context across multiple AI tools and applications without re-uploading or re-configuring.
-
Context Pruning
The practice of removing or masking older, low-signal content, usually stale tool observations, from an AI agent's context window as a trajectory grows.
-
Context Rot
The gradual degradation of an AI system's usefulness as the context it relies on becomes stale, incomplete, or outdated.
-
Context Window
The maximum amount of text (measured in tokens) that a language model can process in a single inference call.
-
Epistemic Provenance
Structural metadata attached to each piece of AI context that records its source, position inside that source, ingestion time, and typed relationships to other context.
-
Fine-Tuning
A training process that adjusts a pretrained language model's weights on a domain-specific dataset to change its default behavior, style, or task performance.
-
Knowledge Graph
A structured representation of information as entities (nodes) and the explicit relationships (edges) between them, queried by traversing connections rather than matching text similarity.
-
MCP (Model Context Protocol)
An open protocol, developed by Anthropic, that standardizes how AI applications provide context and tools to language models.
-
MCP Resources
Read-only, URI-addressable content that MCP servers expose for application-controlled loading, separate from tools that the model invokes autonomously.
-
MCP Server
A server that implements the Model Context Protocol, exposing data and tools to MCP-compatible AI clients.
-
Multi-Agent System
An architecture where multiple AI agents collaborate on a task, each with its own context window, tools, and responsibilities.
-
Prompt Caching
A technique that stores computed key-value tensors from a prompt's prefix so they can be reused on subsequent API calls, reducing cost and latency.
-
Prompt Engineering
The practice of designing the text of prompts sent to a language model to improve the quality, format, or reliability of its output.
-
RAG (Retrieval-Augmented Generation)
A technique that retrieves relevant documents or data at inference time and injects them into the model's context window before generating a response.
-
Semantic Search
A search method that finds results based on meaning and intent rather than exact keyword matching.
-
Structured Context
Context delivered to AI models as organized, typed records with named fields rather than raw prose or unformatted text.
-
Tool Poisoning
An attack in which malicious instructions are embedded in an MCP tool's description or metadata so an agent treats them as authoritative context and acts on them.
-
Wire
A Context as a Service platform that lets you create composable context containers accessible to AI agents through MCP.
Put context into practice
Create your first context container and connect it to your AI tools in minutes.
Create Your First Container