Key concepts
A quick reference for the vocabulary used across Spark and these docs.
Spark uses a small, consistent vocabulary. Once these terms click, the rest of the documentation stays short because it can lean on them. This page is a quick reference, so skim it now and come back whenever a word needs pinning down.
The big picture: a workspace holds your data, which you organize into Spaces, visualize on canvases, act on through tasks, files, and workflows, and hand off to the Agent. Everything below is a piece of that sentence.
Where everything lives
| Term | What it means |
|---|---|
| Organization | The account-level container managed by your identity provider. It holds one or more workspaces. |
| Workspace | Your team's tenant: its own data, members, and settings. The unit of isolation, so data never crosses between workspaces. |
| Space | A container that packages a single use case by pinning the objects, canvases, tasks, workflows, and agents that belong together. |
Working with data
| Term | What it means |
|---|---|
| Data object | A structured collection of records, like a table with typed columns. Your core data lives here. |
| Record | A single row in a data object: one counterparty, one position, one document. |
| Attribute | A typed field on a data object (text, number, date, relationship, and so on): the columns of the table. |
| Relationship | A link between records, often across objects, so related data travels together. |
| View | A saved way of looking at a data object: a specific set of filters, sorting, and grouping. |
Learn more in Data objects (including relationships) and Views.
Seeing and tracking work
| Term | What it means |
|---|---|
| Canvas | A dashboard built from widgets, backed by your data. |
| Widget | A single element on a canvas: a table, chart, or KPI tile. |
| Task | A tracked work item that can be assigned, given a status, and linked to the data it relates to. |
| File | A document or input stored in the workspace, uploadable manually or ingested automatically. |
Learn more in Canvas, Tasks, and the Files section.
Automating operations
| Term | What it means |
|---|---|
| Workflow | An automation that runs a repeatable operation from a trigger through one or more actions. |
| Trigger | The event that starts a workflow: a schedule, a new file, or a change in data. |
| Action | A step a workflow performs once triggered. |
| Human-in-the-loop | An approval step that pauses a workflow for a person to approve or reject before it continues. |
Learn more in Workflows and Human-in-the-loop.
The Agent
| Term | What it means |
|---|---|
| Agent | Spark's agentic layer: it runs inside your workspace, sees your context, and can take real actions on your behalf. |
| Tool | A specific capability the Agent can use to perform an action. |
| Skill | A packaged set of instructions that shapes how the Agent handles a particular kind of task. |
| Connector | A link to an external system the Agent can reach, including over MCP. |
| MCP | Model Context Protocol, a standard way to expose external tools and data to the Agent through a connector. |
| Vault | A secure store for secrets (like API keys) so the Agent can authenticate without exposing credentials. |
Learn more in the Agent section: tools, skills, connectors & MCP, and vaults.
Getting data in and out
| Term | What it means |
|---|---|
| Import | Bringing external data into a data object, for example from a CSV. |
| Export | Taking data out of Spark in a portable format. |
| SFTP ingestion | Automatically pulling files that land on an SFTP endpoint into the workspace. |
Learn more in Add, edit & import records, Manage data views, and Set up SFTP.
Where to go next
- See the concepts in depth: Core concepts.
- Put them to use: Quick start.