# Glossary

> Definitions of the terms used across the APIANT.ai documentation, each linked to the page that explains it.

An automation is the unit APIANT.ai runs: one trigger followed by steps. The terms below are the ones the product screens, Claude and the assistant use. Each entry links to the page that covers the term in full.

## A

**Account.** The login that owns automations, connections and run history, and that a plan applies to. See [Your account](/docs/account).

**Action.** A step that does something in an app or in APIANT, such as creating a record, sending an email or looking up a value. See [Actions and built-in apps](/docs/automations/actions).

**Assembly.** The stored definition behind a connector, an action or a trigger: how APIANT authenticates to the app, which request it sends, and which fields go in and come out. Claude builds and edits assemblies when an app or operation is missing from the catalog. See [Build a connector, action or trigger](/docs/apps/custom-connectors).

**Associated account.** Access one person has to another person's account, for example a contractor working in a customer's account. The associate switches into the owner's account to work in it. See [Linked and associated accounts](/docs/account/linked-accounts).

**Assistant.** The APIANT assistant, the chat panel inside app.apiant.ai that builds, edits and explains automations. It is separate from Claude working over MCP. See [The APIANT assistant](/docs/assistant).

**Assistant credits.** The monthly allowance that pays for assistant messages. A message that starts a model turn costs 20 credits. Working through Claude over MCP does not spend credits. See [Assistant credits](/docs/assistant/credits).

**Automation.** A trigger plus the steps that run each time it fires. An automation is either on (active) or off. See [Automations](/docs/automations).

## C

**Catalog.** The list of apps APIANT.ai can connect to, with their triggers and actions. See [App catalog](/docs/reference/apps).

**Chat widget.** A hosted chat page backed by an automation. Each message a visitor sends starts a run of that automation. See [Chat widgets](/docs/interfaces/chat-widgets).

**Claude Code plugin.** The APIANT skills plus the connection to the APIANT MCP server, installed into a Claude Code project from the account menu. See [Connect Claude Code](/docs/mcp/claude-code).

**Commit.** Saving Claude's draft of an automation back to the automation, which makes it the version that runs. See [Drafts and versions](/docs/automations/versions).

**Condition.** A step that sends a run down one branch or another based on data. Loops and parallel branches are the other control-flow steps. See [Conditions, loops and parallel branches](/docs/automations/control-flow).

**Connection.** Your authorized account in an app (an OAuth sign-in or an API key) that automations use to call that app. See [Connecting accounts](/docs/apps/connections).

**Connector.** The assembly that defines how APIANT signs in to an app. Every connection is made through a connector, and an app's actions and triggers use that connection. See [Apps and connectors](/docs/apps).

## D

**Deploy.** Copying an automation, or a folder of automations, into other accounts, linked or unlinked. See [Publish and deploy](/docs/account/publish-and-deploy).

**Draft.** A working copy of an automation that Claude edits before committing. The committed automation keeps running unchanged until the draft is committed. See [Drafts and versions](/docs/automations/versions).

## E

**Execution history.** The record of an automation's runs, with each step's input, output and errors. How long it is kept depends on your plan. See [Execution history](/docs/runs).

## F

**Fan-out and fan-in.** Starting several child automations at once (fan-out) and pausing the parent until all of them finish (fan-in). The children are counted in a named **latch group**. See [Fan-out and fan-in](/docs/automations/fan-out-fan-in).

**Field mapping.** The expression that fills a step's input field from trigger data or an earlier step's output. See [Field mapping](/docs/automations/field-mapping).

## H

**Hosted form.** A public form page served by APIANT. A form can start a run on every submission, or pause a running automation until one person fills it in. See [Hosted forms](/docs/interfaces/forms).

**Human approval.** A step that pauses a run until a person approves or denies an item from a moderation link, optionally with a timeout. See [Human approval](/docs/interfaces/approvals).

## L

**Linked account.** A child account attached to a parent account. A child account can use the connections its parent chooses to share. See [Linked and associated accounts](/docs/account/linked-accounts).

**Lookup table.** Key-value storage that keeps data between runs. Each saved value has a scope (This automation, My account or Linked accounts) and an expiry (90, 180 or 365 days, or Never). See [Storing data between runs](/docs/automations/data-storage).

## M

**MCP (Model Context Protocol).** The open protocol Claude Code uses to call APIANT. APIANT's MCP server is `https://mcp.apiant.ai/mcp`. See [APIANT over MCP](/docs/mcp).

**MCP tool automation.** An automation that an external AI agent can call as a tool through your account's MCP URL. See Automations as MCP tools.

## O

**Operation.** One trigger or action that an app offers in the catalog.

## P

**Polling trigger.** A trigger that checks an app for new or updated records on an interval. See [Polling triggers](/docs/automations/triggers/polling).

**Publish.** Moving an automation or assembly from a development environment to production. See [Publish and deploy](/docs/account/publish-and-deploy).

## R

**Retry.** A later attempt at a run that failed with a retry-eligible error. APIANT retries on a fixed schedule of up to 9 attempts in total. See [Retries, replays and restarts](/docs/runs/retries-and-replays).

**Run.** One execution of an automation, from the trigger firing to the last step. See [Execution history](/docs/runs).

## S

**Scheduled trigger.** A trigger that fires an automation at set times. See [Scheduled triggers](/docs/automations/triggers/schedules).

**Skill.** A set of instructions in the Claude Code plugin that Claude follows for one kind of work, such as building or testing an automation. You describe the work and Claude picks the skill. See [Claude Code skills](/docs/reference/skills).

**Snooze.** A step that pauses a run until a date and time, then continues. See [Delays, snooze and waits](/docs/automations/delays).

**Step.** One node in an automation after the trigger: an action, a condition, a loop, a subroutine call or another control-flow step. See [Core concepts](/docs/concepts).

**Step comment.** A note you leave on a step in the automation diagram for Claude to apply. See [Review a diagram and send comments to Claude](/docs/mcp/diagram-comments).

**Subroutine.** An automation that another automation calls, passing inputs and receiving outputs. See [Subroutines and calling automations](/docs/automations/subroutines).

## T

**Task.** The usage unit your plan counts each month. A task is counted when an action step that calls an app completes successfully. See [Plans, usage and billing](/docs/account/plans-and-limits).

**Trigger.** The event that starts an automation: a webhook, a poll, a schedule, a form submission, a chat message, a web service call, a call from an AI agent that uses the automation as a tool, or a call from another automation. See [Triggers](/docs/automations/triggers).

**Two-way sync.** A pair of trigger and action settings that keeps records in step between two apps without each update echoing back. See [Two-way sync](/docs/automations/two-way-sync).

## V

**Version.** A saved earlier copy of an automation. Each save keeps the previous copy, so you can ask Claude to list or compare versions. See [Drafts and versions](/docs/automations/versions).

## W

**Web service automation.** An automation behind a URL that holds the HTTP request open and returns the response the automation builds. See [Web service automations](/docs/interfaces/web-services).

**Webhook trigger.** A trigger that fires when another system sends an HTTP request to a URL APIANT generates for the automation. See [Webhook triggers](/docs/automations/triggers/webhooks).

## Next steps

- [Core concepts](/docs/concepts)
- [Public URLs](/docs/reference/urls)
- [Limits](/docs/reference/limits)