AI Context Glossary

AI Agent

An autonomous software program that uses a large language model to plan and execute multi-step tasks.

Unlike a simple chatbot that responds to single prompts, an AI agent can take actions like browsing the web, writing files, and calling APIs to complete a goal. Agents need access to relevant context to do their work effectively, which is why context management systems like Wire exist.

Context as a Service

A model for delivering structured, AI-optimized context to agents and tools on demand.

Instead of embedding context in every prompt or storing it in siloed tools, Context as a Service externalizes it into portable containers. AI agents query the service at runtime, getting exactly the context they need. Wire is a Context as a Service platform.

Context Container

A portable, shareable unit of organized context (documents, data, and structured information) made accessible to AI agents through MCP tools.

A context container in Wire holds your files and structured data, automatically processes them into AI-optimized format, and exposes them via built-in MCP tools (explore, search, write, delete, analyze). Containers are private by default, can be shared with teams, and work across any MCP-compatible AI tool.

Structured Context

Context delivered to AI models as organized, typed records with named fields rather than raw prose or unformatted text.

Structured context improves AI accuracy by reducing token waste, exploiting model attention patterns, and enabling precise retrieval. Research shows format choice alone can swing LLM accuracy by up to 40%. Wire automatically transforms uploaded files into structured, typed records that agents can query efficiently.

Context Engineering

The practice of deliberately designing, structuring, and managing the information provided to AI models to improve output quality and relevance.

Context engineering goes beyond prompt engineering by focusing on what information an AI system has access to, not just how instructions are phrased. It involves decisions about what to include and what to leave out, how to structure it, when to retrieve it, and how to keep it current. Wire automates much of the mechanical work of context engineering.

Context Portability

The ability to use the same context across multiple AI tools and applications without re-uploading or re-configuring.

Context portability means your research notes, company documents, or project files aren't locked into a single AI tool. With Wire, you create a container once and connect it to Claude, Cursor, Cline, or any MCP-compatible tool. Your context follows you.

Context Rot

The gradual degradation of an AI system's usefulness as the context it relies on becomes stale, incomplete, or outdated.

Context rot happens when the information you've provided to an AI tool no longer reflects reality: a product has changed, a team member left, a process was updated. Wire addresses context rot by making it easy to update containers and keep AI context current.

Context Window

The maximum amount of text (measured in tokens) that a language model can process in a single inference call.

Every LLM has a finite context window. Fitting all relevant information into this window is a core challenge in AI development, especially for long documents or large knowledge bases. Techniques like RAG and structured context containers help work around this limitation by retrieving only what's needed.

MCP (Model Context Protocol)

An open protocol that standardizes how AI applications provide context and tools to language models.

MCP, developed by Anthropic, defines a standard way for AI tools (like Claude Desktop or Cursor) to connect to external data sources and capabilities. Instead of every AI tool building custom integrations, MCP provides a common interface. Wire generates MCP servers for every context container automatically.

MCP Server

A server that implements the Model Context Protocol, exposing data and tools to MCP-compatible AI clients.

An MCP server is what AI tools like Claude Desktop connect to in order to access external context. Wire automatically creates an MCP server for each context container, with custom tools tailored to your data. Your AI tools can browse, search, and query your context without any manual configuration.

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.

RAG is commonly used to give LLMs access to knowledge bases that are too large to fit in the context window. Wire uses semantic search and structured retrieval to implement RAG automatically. When an agent queries a container, Wire retrieves the most relevant context and returns it.

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.

Fine-tuning is best suited to changing how a model behaves rather than what it knows. It reliably shifts output format, tone, and narrow task performance, but it struggles to reliably encode new factual knowledge. Research consistently shows RAG outperforms unsupervised fine-tuning for knowledge injection, which is why most production systems fine-tune for behavior and use retrieval for facts.

Semantic Search

A search method that finds results based on meaning and intent rather than exact keyword matching.

Semantic search uses vector embeddings to represent text as mathematical points in space, allowing it to find conceptually related content even when the exact words differ. Wire includes semantic search on all context containers, so agents can find relevant information using natural language queries.

Multi-Agent System

An architecture where multiple AI agents collaborate on a task, each with its own context window, tools, and responsibilities.

Multi-agent systems divide complex work across specialized agents (e.g., a planner, a researcher, a coder) that coordinate through structured handoffs. The main challenge is context management: how agents share information without leaking irrelevant state, duplicating tokens, or operating on stale data. Effective multi-agent architectures scope context per agent and summarize at handoff boundaries.

Context Compression

The practice of reducing token count in an AI agent's context window while preserving the information needed to complete tasks.

As AI agents work through multi-step tasks, they accumulate conversation history, tool outputs, and observations that dilute attention. Context compression techniques like structured summarization, tool response offloading, and embedding-based reduction keep the working context focused. Research shows effective compression can reduce memory usage by 26-54% while preserving task performance.

Context Drift

The gradual loss of task-relevant information as an AI agent's context window fills with accumulated history.

Context drift occurs when new tool outputs, observations, and messages push critical information like the original task goal or early decisions out of the model's effective attention range. Unlike hitting a token limit, drift degrades performance silently. Studies attribute 65% of enterprise agent failures to context drift during multi-step reasoning rather than raw context exhaustion.

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.

Unlike prompt injection, which ends when a session closes, context poisoning persists: the payload is written into sources the agent reads on every future run, such as a vector index, long-term memory store, or shared multi-agent workspace. It is classified as ASI06 in the OWASP Top 10 for Agentic Applications 2026. The root cause is a context engineering failure, since most agent pipelines ingest, store, and retrieve content without provenance tracking, source isolation, or trust boundaries.

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.

AI agents reprocess the same system instructions, tool definitions, and context on every API call. Prompt caching eliminates this redundancy by reusing previously computed representations when the prefix matches exactly. Major providers offer up to 90% discounts on cached input tokens, making it one of the most impactful optimizations for production agent workloads.

Wire

A Context as a Service platform that lets you create composable context containers accessible to AI agents through MCP.

You create containers, upload files or let agents write entries directly, and Wire processes everything into AI-optimized context with built-in MCP tools. Teams use Wire to share context across all their AI tools without duplication or configuration overhead.