Glossary
The words Nagent's screens use, each in a sentence or two, with a link to the page that explains it properly.
Browse documentation
Every word here is one you will meet on a screen
Nagent uses a small vocabulary, and uses it the same way everywhere. This page defines each word as the product means it. Where a word has two meanings, both are given.
A to G
Action class. Which of six kinds an agent's action is: C1 Read, C2 Prepare, C3 Internal act, C4 External act, C5 Publish, or Locked. The class decides the lowest level that runs the action on its own. See autonomy and governance.
AI chief of staff. The agent that heads a department and reports to you: SERA for Sales, MIRA for Marketing and CREA for Content, shown as the AI Sales, AI Marketing and AI Content Chief of Staff. It leads its own room and the team leads under it, and it is the only agent that carries the badge. See prebuilt teams.
AI teammate. An agent that works in your workspace with a role, skills, your knowledge, a memory and an autonomy level. It sits on a team beside your people and is addressed the same way they are. See the Agent Workbench.
AI team lead. An older name for an AI chief of staff. See AI chief of staff.
Approval. A person's yes before an agent's action runs. An action that needs one waits rather than proceeding, and you decide it where the work happened or on the Actions page. See approvals and the inbox.
Autonomy ledger. The Governance page: every agent's trust, karma, errors, blocks and autonomy, banded from healthy to critical. It changes no level. See autonomy and governance.
Autonomy level. How much an agent may do without asking, from L1 to L5: Observe, Assist, Operate, Represent and Lead. Every action belongs to a class, C1 to C5 or Locked, and a level runs the classes up to its own number on its own; anything above waits for approval. A level rises one step at a time, by a person's decision. See autonomy and governance.
Bench. An older word for the agents that work under a team lead, now called its sub-agents. See Sub-agent, and Seat for the places a prebuilt team brings.
Beta. A label on a feature that is new and may be rough in places. A Beta feature is on for everyone; the label tells you what to expect, and nothing marked Beta is held back behind a switch. See huddles for one.
Board. A team's one list of work, shared by people and agents. A task assigned to an agent is picked up without a second press; what the agent then does still faces its own approvals. See the team room.
Brand Lock. Your brand book, on the Brand page, and the switch that sends it with every image and video render. With it on, RUPA and ALPHA work from your palette, logo and glossary. See Brand Lock.
Brief. A short statement of the outcome you want, sent to a chosen agent from the team's agent surface. It becomes a task on the board. HOOK also sends production briefs to ALPHA, which are briefs in the same sense. See sending a brief.
Cap. A spending limit that is checked before money is spent. Each agent has a daily and a monthly cap; meetings and briefs can carry caps of their own. See credits and budgets.
Capture. What Nagent does when a message in a room reads as an instruction: it shows a card, "Instruction read, waiting on you", with Confirm, Edit and Not a task. Only a person's confirmation turns a capture into a task. See sending a brief.
Charter. A team's statement of what it exists to do, kept in
charter.md and shown on the room's Team handbook tab. Any member can
edit it, and the edit is recorded. See
anatomy of a team.
Chief of staff. Short for an AI chief of staff. See AI chief of staff.
Credits. The unit Nagent measures spend in. Your plan carries a monthly allowance of credits, every billable call is recorded against it, and ALPHA quotes video work in the same credits before it starts. See credits and budgets.
Department. One of the areas on the sign-up screen, Sales, Marketing, Content and Customer Experience, each with an AI chief of staff and its team leads. You take a department whole or tick one of its team leads on its own, and the org chart groups your agents by department. See signing up.
Gate. A check that work must pass before it can go further. HOOK's GATE checks ad copy and images against your claims and rules before they ship. The workspace's daily spend limit is checked whenever costly work starts. See HOOK.
Governance profile. The posture your whole workspace takes on autonomy. Standard: agents act within their autonomy level and ask first on anything destructive. BFSI: stricter defaults for regulated work, one level stricter, so more actions wait for a person. See autonomy and governance.
Growth number. The one figure your workspace is run against, set as a metric, a target and a period from a team room's set-up panel. Home tracks it, and where it is a number DRIS measures, DRIS takes it as its objective too. See your first hour.
H to Z
Huddle. A meeting you start yourself, with Huddle at the top of a team room. You say in one line what it should settle and tick who joins; the room's lead runs it, and it ends with decisions and tasks on the board. Its minutes can also add to the team, on the lead's word. See huddles.
Instruction. A message in a room that asks for something to be done, as opposed to a question. A question is answered; an instruction is captured and waits for a person to confirm it. See the team room.
Job. Work a room's lead starts when what you asked needs more than a reply. It runs on its own, with a card in the thread that shows each step, takes Stop, and is steered by a reply. See long-running work.
Knowledge Hub. Where your own material lives for agents to read: documents, pasted text, web pages and your website, organised and searchable. Agents ground their work in it rather than in general knowledge. See the Knowledge Hub and adding knowledge.
Lead. Two meanings. A team's lead is the agent that plans the team's work and hands it out, whether an AI chief of staff or a team lead leading a room of its own. In the CRM, a lead is a person or company you might sell to. See CORA's CRM.
Level ladder. The row at the top of the Agent Workbench that counts your agents at each autonomy level, from L1 Observe to L5 Lead. Press a level to pick out the agents at it. See the Agent Workbench.
Locked. The class of action no level runs on its own: deleting or taking down, redirects, governance changes, lifting a check, spending past a cap, and creating agents, skills, workflows or triggers. It always waits for a person.
Meeting. Two meanings. An agent meeting is your agents talking a problem through within a budget, leaving a transcript, minutes and tasks on the board. A customer meeting is a call or demo with a lead, kept on the lead's Meetings tab. See meetings.
Monid. A paid service Nagent uses to read public data for your agents, such as reviews, social posts and ad libraries. Each call is paid from your credits after its live price is checked, and the endpoints Nagent's own agents read are set up for you the first time one is needed. See third-party tools.
Nomination. An approval card the nightly level run files when an agent has earned a move one level up. A person decides it; nothing is promoted automatically.
Objective. The number DRIS works towards: a metric, a target and a date, declared once the baseline is captured and graded every month as ahead, on target or behind. The set-up chat asks what your objective is before anything else. See DRIS.
Park. A job that is waiting on the platform rather than on a person, such as an ALPHA video job whose render lane is not connected yet. A parked job keeps its place and its approved quote, spends nothing while it waits, and resumes on its own. See ALPHA, the video studio.
Playbook. In HOOK, the playbook is what has worked in your ads, by tag, with its winner rate. See HOOK. A team's Overview lists its workflows under Workflows, not playbooks: the published and draft workflows the team owns or runs, with what starts each. See the team room.
Preset. A ready-made format in RUPA or ALPHA that sets the size or ratio, the length and the kind of shots, such as a Meta Reels ad. No preset works from your description alone. See RUPA and ALPHA.
Promotion review. The button on an agent's page that moves it one level up, after showing its evidence for the last 30 days. A move to L5 needs a second person to approve it only in a workspace with two approvers switched on.
Proposal. A change an agent puts forward that waits for a person before it runs: an agent or team from the builder chat, a fix from DRIS, a budget change from NIA. You approve it where it was made or on the Actions page. See approvals and the Inbox.
Roster. Who is on a team, people and agents, kept in roster.md and
shown as Roster and rank on the team's Overview with each agent's level
and trust. It updates for every member added or taken off. See
anatomy of a team.
Run. One execution of a workflow, or one piece of agent work, from its start to its report. A run started from a room reports back into it, and a produced run can be taken away as a file. See workflows in a room and sending a brief.
Seat. A place on a team for one agent. A prebuilt team names every seat; one marked Coming soon is not built yet and is neither installed nor charged for, and a lead room's set-up list counts whether every seat is filled. See the Agent Workbench.
Specialist. An older name for a team lead. See Team lead.
Skill. A written procedure an agent follows: when it applies, the steps, the rules and what to do when something goes wrong, in plain markdown. A skill attached to an agent is part of its instructions. See skills.
Sub-agent. An agent that works under a team lead on one part of its work, such as HOOK's SCOUT, ANGLE, SCRIPT and GATE, or Echo's REVIEWS, LOCAL and APPS. A sub-agent sits in its team lead's room. See HOOK.
Surface. Where a team's work is laid out rather than talked about. The room's Team surface tab holds the board, the Overview and the lead's page, and a Layout row sets how each kind of board entry is shown. AI Surface is the set of entry types an agent may write. See boards and surface views.
Target. Two meanings in DRIS. The figure an objective aims for, set under The number. And a page DRIS watches, whose prompts are put to the answer engines and listed under Target runs. See DRIS.
Task. One item on a team's board, owned by a person or an agent, and linked to the message or brief it came from. See the team room.
Team. A group of AI teammates and people that owns one outcome, led by a lead agent. A team is either built by Nagent, such as Sales, Marketing and Content, or made by you. See anatomy of a team.
Team lead. Two meanings. As an agent, one that reports to an AI chief of staff and owns one line of work: DRIS for AI visibility, NORA for outbound research, HOOK for ad creative, RUPA for images. A team lead may have sub-agents of its own, and then leads its own room. See prebuilt teams. As a person's role, a Team lead runs one team: see people and roles.
Team room. A team's own space, shown as a channel down the left. It holds the conversation, the agent surface where you brief agents and watch their runs, and the team's files and decisions. See the team room.
Thread. The room narrowed to one exchange, opened with Start a thread. A thread that reaches 40 messages offers to become an agent meeting. See the team room and meetings.
Trigger. What starts a workflow or an agent's action without a person
pressing: a schedule, a business event such as lead.replied, or a signed
webhook from an outside service. A workflow's triggers are off until you
switch them on. See triggers and webhooks,
building a workflow and, for an agent's own triggers,
the Agent Workbench.
Was this right? A card that sends something an agent did on its own back to a person to grade. It blocks nothing; a dismissal counts against the agent's record.
Workflow. A sequence of steps that hands work from one agent to the next. It is built from one sentence, published to a team, and started by a trigger, a test run or a press in the team room. See building a workflow and workflows in a room.
Workspace. Your organisation's own space in Nagent, created when you sign up. Everything you and your agents do lives inside it, and nothing in one workspace is visible to another. See workspace set-up and signing up.