Glossary

Business Summary: Plain-language definitions for Nimbus product terms — use this when an article uses a term you do not recognize.

How this fits in Nimbus

These terms appear across connectors, agents, workstreams, governance, and billing. The relationships below show how the most common terms connect.

flowchart TB
  WS[Workstream]
  Agent[Agent]
  Conn[Connector]
  NTU[NTU]
  HITL[HITL approval]
  Mem[Operating memory]

  Conn --> Agent
  Agent --> WS
  HITL --> WS
  WS --> NTU
  WS --> Mem

Terms

Definitions of terms used across the Nimbus knowledge base. Terms are listed alphabetically.

Agent A governed team of AI specialists with defined roles, topology, and approval rules. An agent turns ambiguous business intent into accountable outcomes. Not the same as a single chatbot.

Artifact A durable output from an agent run — a report, analysis, recommendation, or data product that becomes part of your organization's record.

Audit & History The org-wide log of significant events: agent runs, approvals, exports, and configuration changes. Today, per-workstream Activity tabs and decision traces provide the richest live audit detail; the dedicated Audit & History page shows sample data in demo workspaces until the full org-wide feed ships. See Audit and History.

Checkpoint A point in a workstream where the agent pauses and waits for your sign-off before continuing.

Commit (decision) An approved decision that is released to a downstream system — for example, updating a CRM record or publishing a report.

Committed (work) Governed output that passed HITL and was recorded in the operating ledger — analogous to a Git commit. Appears in Process Board Committed column and decision traces.

Commitment (billing) An optional add-on block of NTU usage included in Team or Business plans beyond the base included pool.

Company Wiki The document and vault layer of the Lifecycle Graph — policies, definitions, onboarding material, and approved documents that agents index and cite. Open from the sidebar Company Wiki item; vault nodes also appear on the graph canvas.

Cluster (wiki hull) An automatically classified group of related wiki documents — visible as themed regions in the lifecycle graph (for example Strategy & Planning). Helps agents retrieve relevant policy context.

Planes (wiki) The tree view of your document repository in the Company Wiki — how uploaded and synced files are organized for browsing and agent context. Product UI may label this planes; docs also use vaults and folder trees for the same structure.

Connector A link between Nimbus and an external system (CRM, ERP, data warehouse, messaging tool, etc.). Connectors let agents read data and, when approved, take action.

Decision trace The record of how an agent reached a conclusion — what evidence it used, which agents contributed, and what approvals were required.

Governance The rules that control how agents operate: who can approve outputs, what spend limits apply, and which connector actions require sign-off. In plain terms: the guardrails that keep AI work accountable to your business.

Domain knowledge framework The structured map of your company's products, markets, and relationships — what the Lifecycle Graph visualizes. (Sometimes called an ontology in technical contexts.)

Chargeback / showback Allocating NTU usage visibility to business units or teams so leaders see who consumed AI work — used for internal cost accountability, not customer invoicing alone.

HITL (human-in-the-loop) A workflow where the agent pauses and waits for a human to review and approve before continuing or releasing an output.

Intelligence connector A connector used for reading and analyzing data. Agents use these to gather context before making recommendations.

Action connector A connector used for writing or executing changes in external systems. These require governance approval before use.

Lifecycle Graph Your organization's unified operating memory — domain context (products, markets, competitors), Company Wiki vaults, agent execution, and governance lineage in one graph. Open from the sidebar as Lifecycle Graph (network or tree canvas) or Company Wiki (document vaults) — all views of the same graph. Live today. See Lifecycle Graph.

Live Canvas A collaborative infinite canvas for structured report boards — executive summaries, scorecards, narratives — with revision history, sharing, and HITL approval banners. Open from Collaboration → Live Canvas, Process Board Reports cells, or workstream Reports mode. See Using Live Canvas.

Omnipresent connector A connector available workspace-wide for common collaboration and file tools (for example Slack, Drive, Notion) — usable across agents without rebinding per workstream.

Onboarding readiness score A 0–100 progress indicator for new workspaces. Connecting data, wiki, team, and first workstream increases the score; target ~80–90 within two weeks.

NTU (Nimbus Token Unit) Nimbus's customer-facing unit for metering agent work. You see NTU quotes before a run starts and usage totals in billing.

Orchestrator The central coordinator within an agent that dispatches specialist agents and manages handoffs.

Operating memory The durable record of governed AI work — runs, deliverables, approvals, connector activity, wiki vaults, and governance states — distinct from ephemeral chat. Visualized in the Lifecycle Graph (network, tree, and wiki views), surfaced in workstream Activity and decision traces, and queryable via Perception. Committed work is written to an operating ledger with traceable lineage.

Perception mode The default chat mode where you describe what you need and Nimbus helps clarify intent before running an agent.

Plan mode A chat mode for governed agent runs that produce a report deliverable. You review and confirm a run plan before execution.

Process Board A cross-workstream kanban spreadsheet under Collaboration showing lifecycle stages — Plan, Run, Reports, Review, Committed — for every workstream and run trail. Used for operational triage and deep-linking into workstreams, runs, and Live Canvas. See Using the Process Board.

Process panel The flowchart tab inside Workstream Studio showing agent logic and sub-tasks for one workstream. Distinct from Process Board, which covers all workstreams. See Workstream Studio and process panel.

Run One execution of an agent on a specific task within a workstream.

Sign-off Your explicit approval at a checkpoint before the agent continues or releases an output.

Sub-agent / Specialist An individual AI role within an agent — for example, a financial analyst agent or a market research agent.

Parallel run Work split across specialists on the same goal, including agentic map-reduce: agents take parts of the job at the same time, then the results are combined inside the workstream.

Workstream A scoped execution workspace where an agent runs on a specific goal. Workstreams have a team, connectors, chat, outputs, and an activity history.

Workspace Your organization's environment in Nimbus — containing workstreams, agents, connectors, and settings. The Workspace tab under Settings shows organization name and workspace ID for support and tenant confirmation. See User settings.

Example scenario

While reading Workstream chat modes, you see Confirm & run and HITL. This glossary defines those terms; the article explains how they appear in your workstream.

FAQs {#faqs}

Q: What is the difference between an agent and a workstream? A: An agent does the work. A workstream is the goal that work belongs to.

Q: Are NTUs the same as API tokens? A: No. NTUs meter governed agent work as a business-facing unit with quotes before runs.

Q: What does HITL mean in practice? A: The agent pauses until a named approver signs off at a checkpoint.

Q: Where do I see past decisions? A: Workstream Activity, Outputs, and — when available — org-wide Audit & History.

Q: What is operating memory? A: The durable record of runs, approvals, and deliverables your organization accumulates over time.