Features

Domains

Domains

Group the metrics, datasets, and sources behind one business area, with the rules that define how it is counted.

A domain is traditionally a business area within a workspace, such as Revenue or Users & Subscriptions. A domain groups metrics, datasets, skills, and data sources that answer one set of questions, and holds the counting rules and notes an agent reads before working in that area.

Domains supply agents with business context that is otherwise undocumented: which table is the source of truth for revenue, which sources must not be read, and how an active user is counted. Members can read domains; workspace owners and admins can create and change them.

Domains can be utilized outside of the traditional sense. Anytime a grouping of context and definitions can be deployed to improve agent accuracy domains are a fit for those tools.

How domains work

What a domain holds

Each domain has a name, the question it answers, a description, a status, counting rules, notes, and links to the objects it covers. Status is one of proposed, active, or archived. New domains start as proposed.

Links and roles

Every link carries a role that defines how the domain relates to the linked object.

Role

Links to

Meaning

Owner

Metric, dataset

The domain defines the object. A metric or dataset has exactly one owning domain.

Related

Any object

The object is tagged to the domain.

Avoid

Data source

Agents working in the domain must not read the source.

A link to a data source can name the specific tables, JSON paths, or MCP tools the domain uses. Archived domains hold no ownership claims: when an active domain claims the same metric or dataset as owner, the archived domain's link is downgraded to related.

Counting rules and notes

Rules are short named statements, such as how a paid account is counted or which time zone defines a day. Notes hold any other context an agent requires. Agents read both before building or answering within the domain.

Build domains

  1. Open the Data Catalog and click Add domain.

  2. Dreambase reads the workspace's existing domains and the contents of each connected database, API, and MCP server, then proposes three to five business areas not yet covered. Select one or enter a custom area.

  3. Review the proposal in two steps: the domain first, then its datasets and metrics. Keep, change, or drop each part, then approve. Each approval is a single save.

Data Catalog pages are read-only. Domains are changed through the workspace chat, MCP, or REST.

Saving merges

A save changes only the fields and links it names. New links are added, links with a new role are updated, and links named in remove_items are removed. Saved links not named in the save are kept. On update, omitted fields keep their values and null clears a field. rules is replaced as a whole.

Use domains with agents

Agents should call get_data_domain_context before building or answering within a business area. The context includes the domain's rules and notes, definitions of owned metrics, related metrics, dataset columns and freshness, readable sources, sources marked avoid, and skill ids. Context that exceeds one response is truncated, and truncated names each section that was cut.

list_metrics, list_datasets, list_connections and list_skills accept a dataDomainId to filter results to one domain, and each returned item lists its linked domains.

Tool reference

list_data_domains

MCP and REST · data-domains:read

{ workspaceId, search?, status?, limit?, offset? }

Lists the workspace's domains in workspace order, each with its id, slug, name, question, status, and link counts by kind and role. Page with offset until nextOffset is absent.

get_data_domain

MCP and REST · data-domains:read + connections:read

{ workspaceId, dataDomainId }

Returns one domain in full: fields, rules, notes, and every link with the target's name and role, including the tables, paths, or tools selected from each source. Long link lists are truncated and flagged with itemsTruncated: true.

get_data_domain_context

MCP and REST · data-domains:read + metrics:read + datasets:read + connections:read

{ workspaceId, dataDomainId }

Returns the business context for one domain. Agents should read it before building or answering in that area, follow its rules, and never read a source marked avoid.

create_data_domain

MCP and REST · data-domains:write

{ workspaceId, name, question?, description?, status?, rules?, notes?, items?, remove_items? }

Creates a domain from the supplied rules, notes, and links. Linked objects must belong to the workspace. Not idempotent: after an ambiguous failure, check list_data_domains before retrying.

update_data_domain

MCP and REST · data-domains:write

{ workspaceId, dataDomainId, name?, question?, description?, status?, rules?, notes?, items?, remove_items? }

Updates the supplied fields. items adds or changes links, remove_items removes them, and all other saved links are kept.

Read a domain's context through the API

curl https://api.dreambase.com/v1/workspaces/<workspace-id>/data-domains/<data-domain-id>/context \
  -H "Authorization: Bearer <access-token>"

REST paths: GET and POST /workspaces/{workspace_id}/data-domains, GET and PATCH /workspaces/{workspace_id}/data-domains/{data_domain_id}, GET /workspaces/{workspace_id}/data-domains/{data_domain_id}/context.

Important boundaries

  • Domains require analytics on for the workspace. Without it, the pages are locked and the tools return an error.

  • Only workspace owners and admins, or an agent pinned to the workspace, can create or change a domain. Members have read access.

  • A metric or dataset has exactly one owning domain.

  • A domain holds at most 50 rules and 300 links per save.

  • Domains are not generated from existing skills. A workspace starts with no domains; they are built through chat or the API.

A domain is traditionally a business area within a workspace, such as Revenue or Users & Subscriptions. A domain groups metrics, datasets, skills, and data sources that answer one set of questions, and holds the counting rules and notes an agent reads before working in that area.

Domains supply agents with business context that is otherwise undocumented: which table is the source of truth for revenue, which sources must not be read, and how an active user is counted. Members can read domains; workspace owners and admins can create and change them.

Domains can be utilized outside of the traditional sense. Anytime a grouping of context and definitions can be deployed to improve agent accuracy domains are a fit for those tools.

How domains work

What a domain holds

Each domain has a name, the question it answers, a description, a status, counting rules, notes, and links to the objects it covers. Status is one of proposed, active, or archived. New domains start as proposed.

Links and roles

Every link carries a role that defines how the domain relates to the linked object.

Role

Links to

Meaning

Owner

Metric, dataset

The domain defines the object. A metric or dataset has exactly one owning domain.

Related

Any object

The object is tagged to the domain.

Avoid

Data source

Agents working in the domain must not read the source.

A link to a data source can name the specific tables, JSON paths, or MCP tools the domain uses. Archived domains hold no ownership claims: when an active domain claims the same metric or dataset as owner, the archived domain's link is downgraded to related.

Counting rules and notes

Rules are short named statements, such as how a paid account is counted or which time zone defines a day. Notes hold any other context an agent requires. Agents read both before building or answering within the domain.

Build domains

  1. Open the Data Catalog and click Add domain.

  2. Dreambase reads the workspace's existing domains and the contents of each connected database, API, and MCP server, then proposes three to five business areas not yet covered. Select one or enter a custom area.

  3. Review the proposal in two steps: the domain first, then its datasets and metrics. Keep, change, or drop each part, then approve. Each approval is a single save.

Data Catalog pages are read-only. Domains are changed through the workspace chat, MCP, or REST.

Saving merges

A save changes only the fields and links it names. New links are added, links with a new role are updated, and links named in remove_items are removed. Saved links not named in the save are kept. On update, omitted fields keep their values and null clears a field. rules is replaced as a whole.

Use domains with agents

Agents should call get_data_domain_context before building or answering within a business area. The context includes the domain's rules and notes, definitions of owned metrics, related metrics, dataset columns and freshness, readable sources, sources marked avoid, and skill ids. Context that exceeds one response is truncated, and truncated names each section that was cut.

list_metrics, list_datasets, list_connections and list_skills accept a dataDomainId to filter results to one domain, and each returned item lists its linked domains.

Tool reference

list_data_domains

MCP and REST · data-domains:read

{ workspaceId, search?, status?, limit?, offset? }

Lists the workspace's domains in workspace order, each with its id, slug, name, question, status, and link counts by kind and role. Page with offset until nextOffset is absent.

get_data_domain

MCP and REST · data-domains:read + connections:read

{ workspaceId, dataDomainId }

Returns one domain in full: fields, rules, notes, and every link with the target's name and role, including the tables, paths, or tools selected from each source. Long link lists are truncated and flagged with itemsTruncated: true.

get_data_domain_context

MCP and REST · data-domains:read + metrics:read + datasets:read + connections:read

{ workspaceId, dataDomainId }

Returns the business context for one domain. Agents should read it before building or answering in that area, follow its rules, and never read a source marked avoid.

create_data_domain

MCP and REST · data-domains:write

{ workspaceId, name, question?, description?, status?, rules?, notes?, items?, remove_items? }

Creates a domain from the supplied rules, notes, and links. Linked objects must belong to the workspace. Not idempotent: after an ambiguous failure, check list_data_domains before retrying.

update_data_domain

MCP and REST · data-domains:write

{ workspaceId, dataDomainId, name?, question?, description?, status?, rules?, notes?, items?, remove_items? }

Updates the supplied fields. items adds or changes links, remove_items removes them, and all other saved links are kept.

Read a domain's context through the API

curl https://api.dreambase.com/v1/workspaces/<workspace-id>/data-domains/<data-domain-id>/context \
  -H "Authorization: Bearer <access-token>"

REST paths: GET and POST /workspaces/{workspace_id}/data-domains, GET and PATCH /workspaces/{workspace_id}/data-domains/{data_domain_id}, GET /workspaces/{workspace_id}/data-domains/{data_domain_id}/context.

Important boundaries

  • Domains require analytics on for the workspace. Without it, the pages are locked and the tools return an error.

  • Only workspace owners and admins, or an agent pinned to the workspace, can create or change a domain. Members have read access.

  • A metric or dataset has exactly one owning domain.

  • A domain holds at most 50 rules and 300 links per save.

  • Domains are not generated from existing skills. A workspace starts with no domains; they are built through chat or the API.

A domain is traditionally a business area within a workspace, such as Revenue or Users & Subscriptions. A domain groups metrics, datasets, skills, and data sources that answer one set of questions, and holds the counting rules and notes an agent reads before working in that area.

Domains supply agents with business context that is otherwise undocumented: which table is the source of truth for revenue, which sources must not be read, and how an active user is counted. Members can read domains; workspace owners and admins can create and change them.

Domains can be utilized outside of the traditional sense. Anytime a grouping of context and definitions can be deployed to improve agent accuracy domains are a fit for those tools.

How domains work

What a domain holds

Each domain has a name, the question it answers, a description, a status, counting rules, notes, and links to the objects it covers. Status is one of proposed, active, or archived. New domains start as proposed.

Links and roles

Every link carries a role that defines how the domain relates to the linked object.

Role

Links to

Meaning

Owner

Metric, dataset

The domain defines the object. A metric or dataset has exactly one owning domain.

Related

Any object

The object is tagged to the domain.

Avoid

Data source

Agents working in the domain must not read the source.

A link to a data source can name the specific tables, JSON paths, or MCP tools the domain uses. Archived domains hold no ownership claims: when an active domain claims the same metric or dataset as owner, the archived domain's link is downgraded to related.

Counting rules and notes

Rules are short named statements, such as how a paid account is counted or which time zone defines a day. Notes hold any other context an agent requires. Agents read both before building or answering within the domain.

Build domains

  1. Open the Data Catalog and click Add domain.

  2. Dreambase reads the workspace's existing domains and the contents of each connected database, API, and MCP server, then proposes three to five business areas not yet covered. Select one or enter a custom area.

  3. Review the proposal in two steps: the domain first, then its datasets and metrics. Keep, change, or drop each part, then approve. Each approval is a single save.

Data Catalog pages are read-only. Domains are changed through the workspace chat, MCP, or REST.

Saving merges

A save changes only the fields and links it names. New links are added, links with a new role are updated, and links named in remove_items are removed. Saved links not named in the save are kept. On update, omitted fields keep their values and null clears a field. rules is replaced as a whole.

Use domains with agents

Agents should call get_data_domain_context before building or answering within a business area. The context includes the domain's rules and notes, definitions of owned metrics, related metrics, dataset columns and freshness, readable sources, sources marked avoid, and skill ids. Context that exceeds one response is truncated, and truncated names each section that was cut.

list_metrics, list_datasets, list_connections and list_skills accept a dataDomainId to filter results to one domain, and each returned item lists its linked domains.

Tool reference

list_data_domains

MCP and REST · data-domains:read

{ workspaceId, search?, status?, limit?, offset? }

Lists the workspace's domains in workspace order, each with its id, slug, name, question, status, and link counts by kind and role. Page with offset until nextOffset is absent.

get_data_domain

MCP and REST · data-domains:read + connections:read

{ workspaceId, dataDomainId }

Returns one domain in full: fields, rules, notes, and every link with the target's name and role, including the tables, paths, or tools selected from each source. Long link lists are truncated and flagged with itemsTruncated: true.

get_data_domain_context

MCP and REST · data-domains:read + metrics:read + datasets:read + connections:read

{ workspaceId, dataDomainId }

Returns the business context for one domain. Agents should read it before building or answering in that area, follow its rules, and never read a source marked avoid.

create_data_domain

MCP and REST · data-domains:write

{ workspaceId, name, question?, description?, status?, rules?, notes?, items?, remove_items? }

Creates a domain from the supplied rules, notes, and links. Linked objects must belong to the workspace. Not idempotent: after an ambiguous failure, check list_data_domains before retrying.

update_data_domain

MCP and REST · data-domains:write

{ workspaceId, dataDomainId, name?, question?, description?, status?, rules?, notes?, items?, remove_items? }

Updates the supplied fields. items adds or changes links, remove_items removes them, and all other saved links are kept.

Read a domain's context through the API

curl https://api.dreambase.com/v1/workspaces/<workspace-id>/data-domains/<data-domain-id>/context \
  -H "Authorization: Bearer <access-token>"

REST paths: GET and POST /workspaces/{workspace_id}/data-domains, GET and PATCH /workspaces/{workspace_id}/data-domains/{data_domain_id}, GET /workspaces/{workspace_id}/data-domains/{data_domain_id}/context.

Important boundaries

  • Domains require analytics on for the workspace. Without it, the pages are locked and the tools return an error.

  • Only workspace owners and admins, or an agent pinned to the workspace, can create or change a domain. Members have read access.

  • A metric or dataset has exactly one owning domain.

  • A domain holds at most 50 rules and 300 links per save.

  • Domains are not generated from existing skills. A workspace starts with no domains; they are built through chat or the API.

Triggers

Metrics

© 2026 Dream, Inc. All rights reserved.