agent protocol compliance evaluation
AI agent protocol compliance evaluation
AI agent protocol compliance evaluation tests whether an agent follows the required message, state and authorization rules of an external protocol.
AI Terminology
Search clear AI terminology definitions, acronyms, related concepts and the reviewed videos that explain them.
Showing 261–280 of 705 terms
Clear filtersagent rate limiting
AI agent rate limiting restricts how many actions an agent may perform within a defined period.
agent reliability
AI agent reliabilityAI agent reliability is the degree to which an agent consistently completes intended work correctly and safely under expected conditions.
agent role boundary
An AI agent role boundary defines the tasks, data, tools and authority that belong to a particular agent role.
agent scope isolation
AI agent scope isolation technically separates each agent's accessible work, tools and data from resources outside its role.
agent self-correction
AI agent self-correctionAI agent self-correction is the process by which an agent uses external feedback to detect a failed attempt, revise its work and try again without immediate human intervention.
agent session isolation
AI agent session isolationAI agent session isolation keeps each agent's conversation state and working context separate from other agents by default.
agent skill registry
An AI agent skill registry is a maintained catalog of reusable, versioned capabilities available to AI agents.
agent software development kit
AI agent SDKagent SDKAn AI agent software development kit provides code libraries and interfaces for building and controlling AI agents.
agent steering
AI agent steering changes an active agent run through bounded guidance or controls without immediately terminating it.
agent task scoping
AI agent task scoping defines the objective, boundaries, context and acceptance conditions for work delegated to an AI agent.
agent tool error
AI agent tool errorAn AI agent tool error occurs when an agent cannot successfully invoke or interpret an external tool because of invalid inputs, unavailable capability, permission failure, or an unexpected response.
agent tool-call latency
AI agent tool-call latencyAI agent tool-call latency is the elapsed time added when an agent invokes, waits for, and processes a tool or external service.
agent undo test
An AI agent undo test asks whether an agent can restore the prior state and how much harm a wrong action could cause.
agent-friendly URL parameter
AI-friendly URL parameterAn agent-friendly URL parameter is a query parameter that requests a compact or structured representation of a page for retrieval by an AI tool.
Agent2Agent Protocol
A2AThe Agent2Agent Protocol is an open communication standard that lets independent AI agents discover capabilities, exchange messages and coordinate tasks.
agentic artificial intelligence benchmark
An agentic AI benchmark evaluates how well an AI system plans, uses tools and completes multi-step tasks.
Agentic Commerce Protocol
ACPThe Agentic Commerce Protocol is a commerce protocol for representing shopping operations and transaction state so AI agents can work with merchants.
agentic Git hosting
Agentic Git hosting is repository hosting designed to coordinate the high-volume branches, commits, reviews and automation created by AI coding agents.
air-gapped artificial intelligence
air-gapped AIAir-gapped AI runs inside an isolated environment with no direct network path to less trusted systems or the public internet.
Knowledge graph
Explore AI concepts
Follow general terms into specialised sub-terms. Select any node to open its definition.