Agents and workflows
Agents and workflows
A2A (Agent-to-Agent)
A2A (agent-to-agent) communication is when one AI agent delegates tasks, shares context, or negotiates outcomes with another agent instead of only talking to the user.
Action Receipt
An action receipt is a plain-language record of what an agent did, when, and with what permissions, like a bank notification, not a server log.
Agent
An agent is an AI system that plans steps, uses tools, and acts across multiple turns to complete a goal, not just answer a question.
Agentic UX
Agentic UX is the design of interfaces for software that plans and acts on a user's behalf, tools, files, APIs, and multi-step workflows, not only text replies.
Autonomy Slider
An autonomy slider lets users set how independently an agent may act, from suggest-only to draft to execute, often per task or surface.
Background Agent
A background agent (or cloud agent) runs a task outside the active chat, on a remote VM, queued worker, or local process, while the user works elsewhere.
Canvas
Canvas is a dedicated side surface, panel, tab, or split view. Where AI output lives as an editable artifact separate from the chat stream.
Claw
In agent tooling, a claw is the reach of an AI orchestrator into your environment: spawning sessions, running commands, editing files, or driving browsers on your behalf.
CLI (Command-Line Interface)
A CLI is a text-based interface where users (or agents) run commands in a terminal instead of clicking through GUI screens.
Composer
The composer is the primary input surface where users assign tasks to AI: type prompts, attach files, pick modes, arm connectors, and send.
Connector
A connector is an integration that lets an AI product read from or write to an external app, email, calendar, Slack, Figma, Stripe, or internal APIs.
Diff / Patch Review
Diff or patch review is the UX pattern where users inspect line-by-line changes an agent proposed, code, copy, config, and accept or reject each hunk before apply.
Fork
Forking in AI chat means branching the conversation or task at a decision point, creating a new path without losing the original thread.
Git
Git is version control software that tracks changes to code and files over time through commits, branches, and merges.
Handoff
Handoff is the structured transfer of context, decisions, and artifacts from one step, or one role, to the next in an AI-assisted workflow.
Harness
A harness is the runtime environment that wraps a model with tools, policies, memory, and orchestration so agents can act inside your product or repo.
Hooks / Automations
Hooks (or automations) are event-triggered agent runs, on save, commit, schedule, or CI signal. That execute predefined skills or prompts without manual chat.
Intent Preview
Intent preview shows what an agent plans to do before it executes. Especially for multi-step or irreversible actions.
Loop Engineering
Loop engineering is designing iterative cycles where an AI observes context, acts, checks results, and repeats until a task completes or a human stops it.
MCP (Model Context Protocol)
MCP (Model Context Protocol) is an open standard for connecting AI assistants to external tools, data sources, and services through a shared protocol.
Multitask
Multitask (in AI tools) means running multiple agent jobs in parallel or background, separate threads, files, or repos, without blocking the main conversation.
Orchestration
Orchestration is how an AI product coordinates multiple models, tools, retrieval steps, and human checkpoints into one coherent run.
Plan Mode vs. Agent Mode
Plan mode is when the AI proposes steps, scope, and files before executing; agent mode is when it runs tools and edits immediately within policy.
Project Rules
Project rules are always-on instructions that shape agent behavior for a repo or workspace, often stored as .cursor/rules, AGENTS.md, or team policy files.
Responsive Salience
Responsive salience is when an AI interface automatically increases or decreases oversight UI, explanations, approvals, transparency, based on task risk and user context.
Sandbox
A sandbox is an isolated execution environment where agents run code, browse, or call APIs with limited blast radius, no production data, no unrestricted network.
Subagent
A subagent is a child agent session spawned by a parent orchestrator to handle a subtask, tests, research, lint fixes, while the parent coordinates the overall goal.
Thread / Session
A thread (or session) is a bounded conversation or task run with its own history, memory scope, and artifacts, distinct from other parallel chats.
Tool Permissions
Tool permissions are granular controls for which external actions an agent may perform, read email vs send email, list files vs delete files.
Tool Use
Tool use is when a model invokes external functions (search, calculators, APIs, code runners) to go beyond plain text generation.
Workflow
An AI workflow is a multi-step process that combines prompts, tools, human review, and structured handoffs to reach an outcome, not a single chat reply.
Worktree
A git worktree is a separate working directory tied to the same repository, often on its own branch, so multiple tasks can run in parallel without stashing changes.