> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.cloudraker.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.cloudraker.com/_mcp/server.

# Key concepts

Each entry below is a short definition. For how to do the thing, follow the link to the page that covers it.

New to CloudRaker? Read [What CloudRaker is](/workspace/get-started/introduction) for the mental model first. Then use this page as a reference when a word is unfamiliar.

## The big picture

<img src="https://fdr-prod-docs-files-public.s3.us-east-1.amazonaws.com/cloudraker.docs.buildwithfern.com/f5c0072cbe797ef4a9f6415941dd241266a6796b1b6465c1fd98c1ab02dea95c/docs/assets/images/guide/get-started/how-it-fits.svg?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Content-Sha256=UNSIGNED-PAYLOAD&X-Amz-Credential=AKIA6KXJSKKNFOCF7G4B%2F20260826%2Fus-east-1%2Fs3%2Faws4_request&X-Amz-Date=20260826T020354Z&X-Amz-Expires=604800&X-Amz-Signature=13737381c307633ed690cb4449eb5206b1f366c18073cc48211c4d3273d87113&X-Amz-SignedHeaders=host&x-amz-checksum-mode=ENABLED&x-id=GetObject" alt="How CloudRaker fits together: space types, playbooks, and actions are set up once in the organization; spaces are created from them; running a playbook in a space produces drafts that a person approves into records, documents, and files" />

#### Organization

Your whole CloudRaker workspace. It holds your people, spaces, definitions, and data. One organization is one tenant. If you belong to more than one, switch between them from the sidebar header. Switching re-scopes and reloads the app, so nothing carries across. See [Organization settings](/workspace/admin/organization-settings).

#### The agent (AI assistant)

The built-in AI assistant. It reads documents, fills forms, and drafts records inside a playbook run. The agent produces **drafts** only. A person always approves. See [Approvals and steering](/workspace/playbooks/approvals-and-steering).

#### Draft

A proposed value, record, or document from the agent or automation that no person has approved yet. A draft's content becomes real data only after a person approves it.

#### Grounding

The pointer from an AI-drafted value back to its exact source location. The source is the file, page, and highlighted region of a PDF, or the timecode in an audio file. Grounding lets you confirm a value against its source before you approve it. See [Search and graph](/workspace/spaces/search-and-graph).

## Spaces

Where the work happens, and the template a space is created from.

#### Space

A workspace that holds the files, records, and people for one body of work. You do your work inside a space. Every space is created from a space type. See [Spaces overview](/workspace/spaces/overview).

#### Space type

The reusable template a space is created from. Admins manage space types. A space type decides which playbooks and actions are available to spaces created from it. See [Space types](/workspace/spaces/space-types).

## Playbooks and runs

The reusable process, and one live execution of it.

#### Playbook

An org-level, reusable definition of a process. It holds a goal, an ordered set of tasks, and the actions the agent may run. A space type makes playbooks available to its spaces. See [Playbooks overview](/workspace/playbooks/overview).

#### Run

One execution of a playbook inside a space. The agent works through the run's tasks. It pauses at approval cards for you to review its drafts. See [Run a playbook](/workspace/playbooks/run-a-playbook).

#### Approval

The human gate inside a run. A person reviews a drafted output and approves or rejects it. Approval turns the agent's drafts into committed data. See [Approvals and steering](/workspace/playbooks/approvals-and-steering).

## Actions

Installable capabilities that run over files.

#### Action

An installable, single-purpose capability that runs over files. Examples: extract data, fill a form, sign, or redact. An action runs on its own from a space's Actions tab, or the agent dispatches it inside a playbook run. See [Actions overview](/workspace/actions/overview).

## Data

The structured data a space collects.

#### Record (data object)

A piece of structured data extracted or entered in a space, for example a single invoice. Approved drafts become records. See [Records](/workspace/data/records).

#### Record type

The org-level definition of a record's shape, which is its set of fields. Every record follows a record type. See [Record types](/workspace/data/record-types).

**Terms from the old product that no longer exist.** If you used an earlier version of CloudRaker, these terms are gone: *My work*, *Inbox*, *Projects* and PRJ numbers, *Building Blocks*, task-type playbooks, and a standalone assistant chat. Work now lives in **spaces**. The assistant appears only inside a **playbook run**.

## Where to go next

#### [What CloudRaker is](/get-started/introduction)

The mental model and the one rule that defines the product.

#### [Quickstart](/get-started/quickstart)

Run your first playbook end to end.

#### [Spaces overview](/spaces/overview)

How a space is organized and what its tabs do.

#### [Playbooks overview](/playbooks/overview)

Understand the processes behind every run.