Glossary
Every TeamMate term in one place. Definitions are alphabetical; each links to the page that covers the concept in depth.
Agent
An agent is an AI assistant you configure to do a specific job — answer questions, run tasks, or operate a process. Each agent has its own instructions, model, knowledge, tools, and channels, and lives in a workspace. Agents are the core building block of TeamMate. See Agent.
Agent Hub
Agent Hub is the in-app name for the marketplace: a catalog of pre-built agent templates and solution packages you can install into your workspace in one step. See Marketplace.
API key
An API key authenticates programmatic access. TeamMate uses API keys both for connecting model providers (so agents can call a model) and for the public APIs that let external systems work with your agents. Keep keys secret and rotate them if exposed. See API keys and Generating API keys.
Capability
A capability is a built-in power you switch on for an agent, such as guardrails, file management, code execution, or trigger and workflow management. Capabilities extend what an agent can do beyond plain conversation. See Capabilities.
Channel
A channel is a place where people reach an agent — the web chat, an embedded widget, Slack, WhatsApp, Telegram, email, or voice. Publishing an agent to a channel makes it available to users on that surface. See Channel.
Embedding
An embedding is a numeric representation of a piece of text that captures its meaning, used to find passages relevant to a question. When you add content to a knowledge base, TeamMate splits it into chunks and stores their embeddings so the agent can retrieve the right passages. See Content details & embeddings.
Guardrail
A guardrail is a rule that constrains an agent’s input or output — for example, blocking certain topics or enforcing a policy. Guardrails keep agent behavior safe and on-brand. An overly strict guardrail can block a response, so review them when an agent goes quiet. See Guardrails.
Interface
An interface is the layout and form through which a user interacts with an agent — the default chat, or a custom form with fields you define. An agent can have more than one interface for different audiences. See Interface.
Knowledge base
A knowledge base is a collection of your content — documents, web pages, transcripts — that an agent can search to answer from your material. Attach one or more knowledge bases to an agent so its answers draw on your sources rather than the model’s general training. See Knowledge base.
Marketplace
The marketplace, shown in the app as Agent Hub, is where you browse and install pre-built agent templates and solution packages. See Marketplace.
MCP
MCP (Model Context Protocol) is an open standard for connecting an agent to external tools and data through an MCP server. Adding an MCP server gives your agents a new set of tools without custom integration code. See MCP servers.
RAG
RAG (retrieval-augmented generation) is the technique behind knowledge bases: the agent retrieves the most relevant passages from your content and includes them as context when generating an answer. This grounds responses in your sources and reduces guessing. See Search & retrieval.
Skill
A skill is a packaged set of instructions and know-how you attach to an agent to teach it a specific procedure, injected into the agent only when it is relevant. Skills let you reuse expertise across agents. See Skill.
Solution package
A solution package is a bundle of related agents that install together with a guided onboarding wizard to cover a complete use case, instead of one agent at a time. See Solution packages.
Task
A task is a unit of work an agent runs — once on demand, or on a recurring schedule. Tasks live under Automation and record each run’s status and output. See Tasks overview.
Tool
A tool is a discrete action an agent can call to do something beyond talking — search a knowledge base, send an email, query a connected app, or run code. Tools come from capabilities, integrations, and MCP servers. See Tool & capability.
Trigger
A trigger is the condition that starts a task or workflow automatically — a schedule, an incoming event, or an external webhook. Triggers let agents act without someone starting them by hand. See Triggers.
Workflow
A workflow is a visual, multi-step process you build from nodes — actions, conditions, and agent steps — that run in sequence to automate work. Workflows live under Automation and keep a per-node execution history. See Workflows overview.
Workspace
A workspace is your isolated environment in TeamMate that holds your agents, knowledge bases, members, credits, and settings. Everything you build belongs to a workspace, and members and billing are scoped to it. See Workspace.