Skip to content

Glossary

This glossary focuses on the product terms that matter most in day-to-day operation.

A

Add Case

The action in Histories that turns a real conversation into a reusable evaluation case in Test Suite.

Agent

The conversational system you configure, test, and publish for a specific job.

Agent Version

A saved and optionally published state of an agent. Draft changes are not the same as a live published version.

API Key

A workspace-scoped credential used to access the agent programmatically.

Attachment

A file the agent should always have in context on every turn.

B

Brand

An identity layer used for published experiences, including visual identity and login configuration.

C

Channel

The delivery surface users actually interact with, such as Web or Slack.

Copilot

The built-in assistant that helps operators design agents, diagnose failures, and suggest fixes.

H

Histories

The operational view of real conversations, feedback, and follow-up actions.

I

Improve

The stakeholder-facing feedback action used to mark a specific reply as needing work.

Instruction

The core behavior rules that define how an agent should act, decide, and respond.

K

Knowledge Base

Connected reference material the agent can search when relevant.

L

Live Test

The in-product place to try an agent before exposing it to real users.

O

Organization

The broader team boundary above any individual workspace.

R

Request Form

A structured form used when the conversation needs to collect specific fields rather than continue as free-form chat.

S

Standard

The checklist-style expectation used to judge whether a case in Test Suite passed or failed.

Test Suite

The evaluation surface where operators store important cases and rerun them against agent changes to catch regressions.

W

Whitelist

An access mode where only specifically allowed users can enter the workspace experience.

Workspace

The operating unit where agents, channels, histories, users, and API keys are managed together.