A
Access Control Rule
A configuration that automatically assigns permissions or agents to users based on criteria like email domain, specific email address, or custom metadata (additional info). See Access Control Rules.Account Manager
A user role that can manage specific organizations and their users without full staff privileges. Account managers see analytics and users only for organizations they are assigned to.Additional Info
Custom metadata stored on a user record, often populated from JWT claims during widget authentication. Used by access control rules for dynamic permission assignment.Agent
An AI-powered assistant configured with knowledge sources, tools, and behavior settings. Agents process user messages and generate responses. See Agents.Agent-as-Tool
A feature allowing one agent to invoke another agent as a tool, enabling workflows where a general agent delegates to domain experts. See Agent Invocation Tool.API Key
A secret token used for authenticating API requests. Generated in the admin interface and included in request headers. See API Key Authentication.B
Background Agent
An agent type that processes requests asynchronously. Suitable for long-running tasks like report generation or complex analysis.C
Chunk
A segment of a document created during knowledge source processing. Documents are split into chunks for efficient search and retrieval.Connector
An integration that syncs data between AGO and external platforms. Types include:- Knowledge Connectors: Import documentation (Notion, Confluence, Zendesk Guide, etc.)
- Ticketing Connectors: Sync tickets (Zendesk, HelpScout, HubSpot)
- Communication Connectors: Handle messages (Slack, WhatsApp, Messenger)
Conversation Starters
Clickable buttons displayed on the chat home page that let users start a conversation with a single click. Each starter sends a pre-defined message, reducing friction and guiding users toward common topics. See Conversation Starters.CX Score
A 0–100 quality metric calculated by the Thread Evaluator after each conversation. Measures five dimensions: resolution quality, response coherence, tone, efficiency, and customer effort. Visible in the conversation sidebar and on the performance dashboard. See CX Score.D
Display Name
The user-facing name shown for an agent in the chat interface, as opposed to the internal name used for identification and API calls.Document Type
A category assigned to knowledge documents that determines which quality rules apply. Aligned with documentation frameworks (e.g., tutorial, how-to, reference, explanation). Used by the Knowledge Quality system to evaluate content. See Knowledge Quality.E
Embedding
A way of representing text as numbers so that similar content can be found by meaning rather than exact keywords. AGO converts documents and queries into embeddings to power semantic search.H
Home Page
The landing page users see when opening the chat widget or knowledge portal. Admins can customize it with conversation starters, quick links, and curated resources. See Home Page Management.I
Invisible Router
A transparent routing system that directs conversations to specialized agents without the user seeing a handoff. The routing agent picks the best sub-agent based on the user’s question, and the conversation continues as if the user were talking directly to that agent. See Agents.J
JWT (JSON Web Token)
A compact, URL-safe token format used for authentication. AGO supports JWT-based widget authentication where user claims can be extracted for access control. See Widget Authentication.K
Knowledge Source
A connected repository of content that agents can search and reference. Examples include Notion workspaces, Confluence spaces, Zendesk help centers, and uploaded files. See Knowledge Connectors.L
LLM (Large Language Model)
The AI model that generates agent responses. AGO supports multiple LLM providers and allows per-agent model configuration.M
MCP (Model Context Protocol)
A standardized protocol that lets external systems provide tools and capabilities to AGO agents. MCP connections are tracked with dedicated analytics (call metrics, errors, duration). See MCP Tool API.O
Organization
A tenant entity in AGO’s multi-tenant system. Organizations have isolated data and can be linked to users via email domains. See Organization Management.P
Permission
A named group that controls access to agents, tools, or features. Users are assigned to permission groups, and agents are made visible to specific groups.Pinned Document
A document that is always included in an agent’s context, regardless of how relevant it is to the current query. Useful for critical policies or FAQs.Pre-chat Form
A form displayed before the chat begins that collects user information or qualifies users by asking a question with answer options. Each answer option can assign a permission that controls which agents the user sees. See Pre-chat Form.Prompt Template
A reusable set of instructions that controls an agent’s behavior, tone, and response format. Prompt templates can be versioned and shared across agents. See Prompt Templates.R
RAG (Retrieval-Augmented Generation)
The process of searching your knowledge base for relevant documents before generating a response. This grounds the agent’s answers in your actual content rather than general knowledge.Reasoning Agent
An agent type with extended reasoning capabilities that performs multiple reasoning steps for complex, multi-part problems.Reranker
A secondary model that re-orders retrieved documents by relevance after the initial search. Improves response accuracy by surfacing the most relevant content first.S
Satisfaction Feedback
A feature that collects user feedback (thumbs up/down or smiley ratings) after agent responses. Negative feedback can trigger actions like ticket creation. See Feedback Analytics.SDK (Software Development Kit)
The AGO JavaScript SDK (@useago/sdk) for building custom chat interfaces. Provides client libraries, React hooks, and pre-built components. See JavaScript SDK.
Search Lab
A debugging tool in AI Studio that lets you test document retrieval for specific queries. Shows which documents and chunks the agent would find, helping you diagnose relevance issues.Source Attribution
Citations shown alongside agent responses that identify which documents contributed to the answer. Lets users verify information against the original source.SSE (Server-Sent Events)
A web standard for streaming data from server to client. AGO uses SSE to stream agent responses in real-time so users see text as it is generated.Staff User
A user with administrative privileges who can manage agents, knowledge sources, organizations, and system configuration.Suggested Actions
AI-generated clickable reply options shown during a conversation. Instead of a long answer, the agent presents 2–4 options the user can click, plus a text input for custom replies. See Suggested Actions.T
Temperature
An LLM parameter controlling response randomness. Lower values (0.0–0.3) produce consistent, deterministic responses; higher values (0.7–1.0) allow more variation.Thread
A conversation between a user and an agent, containing multiple messages. Threads maintain context across the interaction.Thread Evaluator
An automated system that analyzes completed conversations and generates a CX Score with dimensional breakdowns, review flags, and improvement suggestions. Runs daily to assess conversation quality. See Thread Evaluator.Tool
An action that agents can perform during a conversation — creating tickets, calling APIs, sending emails, searching with specific parameters, or invoking other agents. See Tools.Topic Analyzer
An AI-powered system that automatically assigns topics to conversations, enabling filtering and analytics by subject area without manual labeling. Previously called Thread Tagger. See Topic Analyzer.Trigger
An automation rule that executes actions when conditions are met:- Ticket Triggers: fire on ticket creation or update
- Thread Triggers: fire on conversation events
U
User Memory
Cross-session memory that stores facts about users (preferences, past issues, product context) extracted from conversations. Agents use this to provide personalized support without requiring users to repeat information. See Agent Memory.V
Viewer
A read-only admin role. Viewers can see conversations, dashboards, and documents, but cannot change configuration or settings.W
Widget
An embeddable chat interface for websites and applications. Configured via a JavaScript snippet and supports customization of colors, layout, and behavior. See Widget Configuration.Webhook
An HTTP callback that AGO receives from external services. Used for integrations like Slack (message events) and Zendesk (ticket events).Related Documentation
- Platform Overview — AGO architecture and components
- Quickstart — get started in 5 minutes
- Security Overview — authentication methods