Skip to content
Browse the docs

Glossary

The words Nested uses — agent, pack, skill, formula, workspace, task, automation, connector, memory, team — and how they fit together.

Updated 3 min readView as Markdown

Nested uses a small set of words consistently across the app, the console, and these docs. Learn them once and every screen reads the same way.

The short version#

A team owns agents. Each agent is a deployment that carries packs and secrets. You give an agent work as a task inside a workspace. Automations start tasks for you. Connectors, knowledge centers, and your memories give agents what they need to know and reach.

Terms#

TermWhat it means
AgentA specialized worker with its own machine, tools, and sign-ins. You deploy it once and give it tasks.
DeploymentThe running form of an agent in your Kubernetes cluster: a cloud-agent pod plus its configuration.
PackA versioned bundle of capabilities you attach to an agent: skills, subagents, formulas, MCP servers, setup scripts, and guides.
SkillA focused instruction set (a SKILL.md file) that teaches the agent how to do one kind of job.
SubagentA specialist the agent can hand part of a job to, defined inside a pack.
MCP serverA tool server the agent talks to over the Model Context Protocol, declared in a pack.
FormulaA reusable, typed workflow: a graph of steps and gates with checked input. Runs open on a run canvas.
WorkspaceA named topic for an agent’s tasks. It is an open folder (a scratch space) or one repository.
TaskOne unit of agent work: a conversation thread with a status. Many tasks run in parallel.
SubtaskA task started from another task. It appears on a lane under its parent.
AutomationA named trigger plus instructions. When it fires — on a schedule, a webhook, GitHub, Slack, or another source — Nested starts a task.
ConnectorAn account-level link to an outside tool (for example GitHub) that agents can use.
SecretA stored credential, such as an API key, that you bind to an agent instead of pasting it into chat.
Knowledge centerA shared, versioned set of Markdown pages that you attach to agents as reference or as a place to write.
MemoryWhat agents remember about you and your work. It belongs to you, not to one agent, and you choose it per task.
TeamThe group that owns agents, workspaces, and tasks. Every account works inside a team.
EnvironmentA short-lived machine an agent creates to run or test an app outside its own container. It runs on a cloud provider, not in your cluster, and only when your team enables Environments.
Site loginA saved sign-in for a website, which an agent can use to log in for you.

How the pieces relate#

Agents, packs, and secrets#

An agent is only as capable as what you attach to it. A pack adds skills, subagents, formulas, and MCP servers. Secrets supply the credentials those packs declare. You pick both in the deploy wizard and change them later in agent settings.

Workspaces and tasks#

A workspace groups related tasks for one agent so topics do not mix. A task is the actual conversation. You pick the agent and the workspace in the composer when you start a task. See Chat, workspaces, and tasks.

Knowledge and memory#

A knowledge center is shared and attached to agents. A memory is yours and is chosen per task. Both use the same page format, with full version history. See Shared knowledge centers and Memories.

Automations and formulas#

An automation decides when work starts. A formula decides how multi-step work proceeds once it starts. An automation can run either plain instructions or a formula from the agent’s packs. See Automations and Formula runs in chat.

Example#

Your team deploys a reviewer agent with a code-review pack and a GitHub token secret. An automation fires when a pull request opens and starts a task in the “Pull requests” workspace. The agent reads the team’s “Coding standards” knowledge center, remembers from your memory that you prefer short comments, and posts its review.

Questions

What is the difference between an agent and a deployment?
They are the same thing seen from two sides. People talk about the agent; the cluster runs it as a deployment.
What is the difference between knowledge and memory in Nested?
A knowledge center is shared and attached to agents. A memory belongs to you and follows you to whichever agent works on your task.
What is a pack?
A pack is a versioned bundle of skills, subagents, formulas, MCP servers, and setup that you attach to an agent. You can build your own in the pack builder.