> For the complete documentation index, see [llms.txt](https://docs.soda.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.soda.io/soda-documentation/soda-ai/soda-mcp/mcp-tools.md).

# MCP tools

Soda MCP exposes Soda Cloud operations as Model Context Protocol tools. This catalog describes the 110 tools in Soda MCP 0.11.0 and covers both [remote](https://docs.soda.io/soda-documentation/soda-ai/soda-mcp/connect-remotely.md) and [local](https://docs.soda.io/soda-documentation/soda-ai/soda-mcp/run-locally.md) connections.

Tools marked **Local only** are not available remotely: `create_secret`, `update_secret`, and `create_runner`. Tool use is subject to your Soda Cloud permissions and organization features.

Most tools map to documented operations within Soda v4. The contract and data-standard knowledge tools provide packaged authoring guidance without calling Soda Cloud.

Use warehouse-enforced read-only grants for ordinary source connections. Soda MCP does not verify or enforce those grants; tool safety hints assume they are in place. Diagnostics destinations need their own write privileges.

<details>

<summary>Attributes - labels for organizing datasets, checks, and columns</summary>

Attributes categorize and organize resources, such as marking checks as critical for notification rules.

* `list_attributes` — List attribute definitions available in your Soda Cloud organization for datasets, checks, columns, data standards, and data products.
* `create_attribute` — Create a new attribute definition for datasets, checks, columns, data standards, or data products.
* `update_attribute` — Update an attribute's label, description, or allowed values.
* `delete_attribute` — Delete an attribute definition. The attribute is removed from any resources where it is set.

</details>

<details>

<summary>Domains - organizational groupings identified by a slug</summary>

A domain is a flat organizational grouping, such as a business unit or team, with a display label and an immutable lowercase slug (`name`). Data product YAML refers to a domain by its `name`; domains hold no datasets or checks. These tools require the `domainsEnabled` organization feature, and writes require **Manage organization settings** permission.

Look up a slug with `list_domains(name=...)` before creating a domain. Before updating, read `get_domain` and preserve the current description unless you intend to change it: `update_domain` replaces label and description together, so omitting the description clears it. A domain referenced by a data product cannot be deleted. Deletion is permanent and asynchronous; an acknowledgment confirms scheduling, not completion.

* `list_domains` — List the domains of your Soda Cloud organization, sorted by `name`. A domain is a flat organizational grouping, such as a business unit or team, with a display `label` and an immutable lowercase slug `name` (for example `finance-analytics`). A data product refers to a domain by the domain's `name`, as the `domain` value of its YAML.
* `get_domain` — Retrieve one domain by `domain_id`. Returns `{domain}` with its `name` (slug), `label`, `description`, `created`, and `lastUpdated`.
* `create_domain` — Create a domain: a flat organizational grouping with a display `label` and an immutable lowercase slug `name` (for example `finance-analytics`). A data product refers to a domain by the domain's `name`, as the `domain` value of its YAML.
* `update_domain` — Replace a domain's `label` and `description`; the slug `name` cannot change.
* `delete_domain` — Delete a domain by `domain_id`.

</details>

<details>

<summary>Checks - data quality validations from multiple sources</summary>

Checks are Soda Cloud data quality validations and can originate from Soda products, dbt, data contracts, data standards, or metric monitoring. Contract-authored checks can use fixed thresholds; metric-monitoring checks use dynamic thresholds. There is no generic public create-check endpoint.

Check discovery and check history exclude monitors. Use `list_monitors` to find a dataset's monitors and `list_monitor_results` for their history. For the latest results of known IDs, `list_checks(check_ids=[...])` accepts 1–100 check or monitor IDs; do not combine that batch with filters, pagination, or sorting. Ordinary SodaCL anomaly checks remain checks.

* `list_checks` — List checks in your Soda Cloud organization, including current status and latest result, associated datasets, agreements, and linked incidents.
* `list_check_results` — List historical results for one check, including result IDs, evaluation statuses, and data timestamps.
* `delete_check` — Delete a specific check by its ID. Checks are the general Soda Cloud validation resource and can originate from contracts, data standards, Soda products, dbt, or metric monitoring.

</details>

<details>

<summary>Datasets - tables tracked in Soda Cloud</summary>

Use `set_column_attributes` to change only the named column attributes, preserving others. Set an attribute to `null` to clear its value; column and attribute names must still be valid.

* `list_datasets` — List datasets in your Soda Cloud organization, including data source, incidents, attributes, health status, and contract metadata.
* `get_dataset` — Get a single dataset by dataset ID.
* `get_dataset_by_dataset_qualified_name` — Get a single dataset by dataset qualified name (DQN), e.g. \[datasource\_name]/\[database\_name]/\[schema\_name]/\[dataset\_name].
* `update_dataset` — Update dataset properties (label, tags, attributes, owners, profiling, metric monitoring, diagnostics warehouse, compute warehouse, time partition, data delay). For tags, owners and attributes, each supplied field replaces its entire value: tags replaces the tag list, owners the owner list, and attributes the attribute mapping. Entries excluded from that supplied value are removed; fields omitted from the update remain unchanged. To add, remove or change individual entries in these fields, use a completed read of the same dataset that explicitly includes the field being edited. Preserve its other entries in the outgoing value. If the read omits that field, its current value is unknown—not empty. Obtain usable state or stop without updating that field. The tool does not fetch or merge these three fields automatically. An explicitly supplied complete final value for one of these fields can replace directly without a preservation read. For a targeted owner edit, project each retained user's user.userId to {type: user, userId: ...} or group's userGroup.userGroupId to {type: userGroup, userGroupId: ...}, omitting response-only metadata; never invent IDs. Omitted/None collections stay unchanged; tags=\[] and attributes={} clear them. Do not call update\_dataset with owners=\[]; explain that removing all owners is unsupported.
* `list_dataset_columns` — List active columns of a dataset with current assigned column attribute values.
* `set_column_attributes` — Set attribute values on columns of a dataset.
* `delete_dataset` — Delete a dataset. This action is permanent and cannot be undone.
* `list_dataset_roles` — List available dataset roles. Use dataset roles to manage access to individual datasets.
* `create_dataset_role` — Create a custom dataset role.
* `update_dataset_role` — Update the name or permissions of a custom dataset role.
* `delete_dataset_role` — Delete a custom dataset role. This action is permanent and cannot be undone.
* `get_dataset_compute_warehouse` — Get the compute warehouse configuration for a dataset.
* `get_dataset_diagnostics_warehouse` — Get diagnostics warehouse information for a dataset.
* `get_dataset_profiling` — Get profiling information for a dataset (structure, statistical summaries, data characteristics).
* `list_dataset_responsibilities` — List user and user group permissions assigned to a dataset, and their associated roles.
* `update_dataset_responsibilities` — Update user/group permissions and their associated roles for a dataset.

</details>

<details>

<summary>Monitors - ML-based dynamic-threshold metrics</summary>

Monitors track metrics over time with anomaly detection. Dataset monitors cover built-in metrics, while column and custom SQL monitors are user-defined.

General monitor management requires a supported V4 dataset and **Manage contracts** permission; creating monitors also requires the relevant monitoring features and dataset setup. Dataset monitors can be disabled with `update_monitor`, but cannot be deleted.

For grouped monitors, read history using the group member IDs; group parents have no results of their own. Custom SQL must return the numeric `resultMetric` label and all grouping labels; aggregate queries must include those labels in `GROUP BY`. Clearing SQL grouping does not rewrite the query: ungrouped SQL must return exactly one metric row. Record-level Anomaly Detection (RAD) requires feature and runner support. Its diagnostics describe partition-level distribution changes, not individual anomalous rows or proven business causes.

Feedback records your judgment about a result's classification. Pass 1–100 result IDs, not monitor or check IDs, to `submit_monitor_result_feedback`. Investigating a result does not submit feedback; confirm any judgment proposed by your agent before it submits feedback.

{% hint style="warning" %}
Creating an enabled freshness, rows-inserted, or RAD dataset monitor can automatically backfill historical data. Completion can delete results not reproduced by the backfill and replace configuration-change history of other included monitors. If that broader history loss is outside your authorized scope, the agent must explain it and ask before creation.

Re-enabling freshness, rows-inserted, or RAD monitoring can reset other monitors' history and start collecting historical data again. Clearing a disabled monitor's `isEnabled` override with `null` can also re-enable it.

Only after you authorize this broader reset, pass `allow_dataset_history_reset=true` as a separate argument to `update_dataset` or `update_monitor`, not inside `body`. Otherwise, affected edits return `dataset_history_reset_required` before any write. SQL query, result-column, or grouping changes clear affected monitor and child results without compensating backfill. Let automatic history collection finish rather than starting it again; starting collection does not mean training is complete.
{% endhint %}

* `list_monitors` — List a V4 dataset's metric monitors and latest outcomes, across dataset, column and custom SQL types.
* `list_monitor_results` — List one metric monitor's historical results and available diagnostics, including RAD explanations.
* `submit_monitor_result_feedback` — Replace classification feedback on 1-100 V4 metric-monitor results with the user's judgment.
* `get_dataset_metric_monitoring` — Get metric monitoring configuration for a dataset, including all three monitor types and the dataset-level scan schedule.
* `create_monitor` — Create a V4 dataset, built-in column, or custom SQL metric monitor using ML-based dynamic thresholds.
* `update_monitor` — Update a metric monitor by ID; the backend resolves its dataset and family.
* `delete_monitor` — Delete a column or custom SQL metric monitor, permanently removing its results and group children.
* `run_historical_metric_collection` — Trigger a historical metric collection scan for a dataset. The v4 API calls this a historical metric collection scan; in user-facing terms it backfills metric history so monitors (ML-based anomaly detection) have a baseline to learn from.

</details>

<details>

<summary>Contracts - per-dataset sources of truth for contract-authored checks</summary>

A data contract defines its contract-authored checks for one dataset. See Editing contracts for the safe edit workflow.

* `list_contracts` — List data contracts in your Soda Cloud organization. A data contract is the declarative source-of-truth for contract-authored checks on a dataset.
* `create_contract` — Create and publish a data contract on a dataset from the YAML `contents` you supply.
* `get_contract` — Retrieve a specific data contract by ID, including its YAML content, with an optional contract-editing mode.
* `edit_contract` — Edit an existing data contract's YAML using its contract ID and the actual `edit_revision` from a completed `get_contract(for_editing=true)` result. This is not a general editor for check YAML: data-standard YAML, including threshold patches and previews, uses `update_data_standard`. Never substitute a data-standard ID or invent a revision. Operations-mode diffs compare the normalized baseline with the patched candidate; contents-mode diffs compare the exact fetched and caller-supplied strings.
* `get_contract_editing_knowledge` — Get packaged contract-editing knowledge, returning the authoring guide by default.
* `list_contract_versions` — List published versions of a specific data contract. Each publish creates a new version; this tool returns the version history.
* `verify_contract` — Trigger a contract verification scan for the given contract. Runs the checks defined in the contract against the dataset.
* `create_skeleton_contract` — Trigger async generation of a skeleton contract for a dataset, derived from its warehouse schema. Use this to bootstrap a contract's YAML body when no contract has been published yet.
* `get_skeleton_contract_status` — Get the status of a skeleton contract generation operation started via `create_skeleton_contract`.
* `generate_contracts` — Trigger async AI-powered generation of full contracts for one or more datasets.
* `get_contract_generation_status` — Get the status of a contract generation operation started via `generate_contracts`.

</details>

<details>

<summary>Data Standards - organization-wide YAML policies applied by scope</summary>

A data standard applies checks to every dataset matching its scope, unlike a contract, which belongs to one dataset. Targeted YAML editing preserves supported public configuration and validates the whole candidate locally. See Editing data standards for modes, preview, and publication boundaries.

Datasets matching any `scope.excludeRules` are excluded even when an inclusion rule matches. Preview candidate matches with `preview_data_standard_scope`. Configuration updates replace the scope as a whole: read the current definition and preserve its scope, including existing exclusions, unless you intend to change them. Omitting or emptying the scope adds no user-defined scope restriction.

* `list_data_standards` — List data standards in your Soda Cloud organization. A data standard is an organization-wide YAML policy whose checks are automatically applied to every dataset matching its scope.
* `get_data_standards_activity` — Get an organization-wide rollup of data standards activity, including counts of active and total standards, matched datasets, and aggregated check results.
* `get_data_standard` — Retrieve a specific data standard by ID, including its YAML contents, attributes, and scope.
* `get_data_standard_authoring_knowledge` — Retrieve the authoritative data-standard authoring guide and YAML JSON schema for creating, editing, and previewing standards. For requests for this guide or schema, call this tool directly; no resource lookup is required.
* `list_data_standard_checks` — List the aggregated checks generated by a specific data standard across the datasets it matches.
* `list_data_standard_datasets` — List the datasets currently matched by a specific data standard's scope.
* `evaluate_data_standard_scope` — Re-evaluate which datasets currently match a data standard's scope and return their dataset IDs.
* `preview_data_standard_scope` — Preview which datasets a candidate data standard scope matches while leaving saved standards unchanged.
* `create_data_standard` — Create an organization-wide data standard whose YAML checks apply to every dataset matching its scope.
* `update_data_standard` — Edit a data standard's YAML checks or configuration by `data_standard_id`.
* `update_data_standard_status` — Transition a data standard to a new status without modifying its contents, attributes, scope, owners, or schedule. Use this to pause or activate an existing standard.
* `delete_data_standard` — Delete a data standard.
* `execute_data_standards` — Trigger a scan that runs the active data standards linked to a dataset. Identify the dataset by `datasetId`.
* `dry_run_data_standard` — Test an unsaved data standard (a dry run) by running its checks once against one dataset's live data, without creating or modifying a public data standard. Provide the standard YAML in `contents` and the target dataset in `datasetId`.

</details>

<details>

<summary>Data Products - input datasets grouped behind a report or dashboard</summary>

A data product groups the input datasets behind a report, dashboard, or other deliverable and reports their combined health. It is defined in YAML and addressed by its `id`, not its name or display label. The YAML can also declare a `domain` (the domain’s slug) and `outputs` (datasets and dashboards). It owns no checks: contracts and data standards define them.

These tools require the `dataProductsEnabled` organization feature and the relevant data product permissions.

Resolve an exact YAML `key` with `list_data_products(name=...)` and use the returned `id` as `data_product_id`. Use `search` for display names instead. If an exact-name lookup finds no product you can view, report that and change nothing; do not substitute a near match before publishing or deleting.

Create and publish accept the complete YAML in `body.contents`. Before publishing, read `get_data_product` and preserve unchanged fields: publishing replaces the definition, so omitted inputs, outputs, domain, attributes, or descriptions are removed. Use `list_attributes(resource_type="dataProduct")` to resolve attribute definitions, and review any warnings returned by create or publish. Deleting a product leaves its input datasets, contracts, and checks unchanged.

Filter products by domain slug with `domain`, or by output types with `output_type`. Inputs and outputs have separate paginated listings; input dataset details are nested under `dataset`.

* `list_data_products` — List data products in your Soda Cloud organization. A data product groups the input datasets behind one report, dashboard, or other deliverable and reports their combined health.
* `get_data_product` — Retrieve one data product by `data_product_id`, including its YAML `contents`, owner, attributes, and health. Its `domain` (`id`, `name`, `label`) is absent when the product has none.
* `list_data_product_inputs` — List the input datasets of one data product, with each input's check-result counts and last evaluation time.
* `list_data_product_outputs` — List the outputs of one data product: the datasets and dashboards that it provides, as declared in its YAML `outputs`.
* `create_data_product` — Create a data product from its YAML definition in `body.contents`.
* `publish_data_product` — Publish new YAML for an existing data product, addressed by `data_product_id`. `data_product_id` is the `id` from `list_data_products`, never the product's `name`.
* `delete_data_product` — Delete a data product by `data_product_id`.

</details>

<details>

<summary>Datasources - warehouse connections that own datasets</summary>

A datasource is a configured warehouse connection, such as Snowflake or Postgres, that owns datasets.

Reading datasource configuration YAML can return stored literal credentials to the MCP host; those values enter model context when the host or embedder forwards the tool result. Soda secret references remain `${secret.NAME}` placeholders. Prefer Soda secrets and the Soda Cloud Web UI for future credential changes. This server does not redact configuration-read results or control host or model-provider retention.

* `list_datasources` — List datasources in your Soda Cloud organization. A datasource is a configured connection to a warehouse (e.g. Snowflake, Postgres) that owns datasets.
* `run_datasource_discovery` — Trigger a discovery scan on a datasource. Detects new tables and schemas in the warehouse on-demand (instead of waiting for the scheduled cron).
* `create_datasource` — Create a new datasource from a YAML configuration document. The datasource type (e.g. Snowflake, Postgres) is extracted from the YAML contents.
* `get_datasource` — Get datasource metadata by ID: ID, name, label, type, owner, and creation/update timestamps.
* `get_datasource_configuration_file_contents` — Get the current datasource connection configuration YAML by ID. Use this read before a targeted edit: supplying `configurationFileContents` to `update_datasource` replaces the whole YAML, so preserve every unchanged returned value.
* `update_datasource` — Update a datasource. Top-level fields are PATCH-like: omitted fields keep their existing values. Supplying `configurationFileContents` instead replaces the entire YAML; for a targeted YAML edit, first call `get_datasource_configuration_file_contents` and preserve every unchanged returned value, including existing literals.
* `delete_datasource` — Delete a datasource and all its associated resources (datasets, checks, scans, incidents).
* `list_datasource_roles` — List available datasource roles. Use datasource roles to manage access to individual datasources.
* `create_datasource_role` — Create a custom datasource role.
* `update_datasource_role` — Update the name or permissions of a custom datasource role.
* `delete_datasource_role` — Delete a custom datasource role. This action is permanent and cannot be undone.
* `get_datasource_diagnostics_warehouse` — Get the diagnostics warehouse configuration for a datasource. The diagnostics warehouse collects scan-related data (failed rows, scan results) and forwards it to the customer's warehouse for storage and analysis.
* `update_datasource_diagnostics_warehouse` — Update the diagnostics warehouse configuration for a datasource. The diagnostics warehouse collects scan-related data (failed rows, scan results) and stores it in your warehouse for analysis.
* `list_datasource_responsibilities` — List user and user group permissions assigned to a datasource, and their associated roles.
* `update_datasource_responsibilities` — Update user/group permissions and their associated roles for a datasource.
* `test_datasource_connection` — Trigger an async connection test for a datasource configuration. Use this to validate a YAML + runner combination without creating a datasource, e.g. before calling `create_datasource`.
* `get_datasource_connection_test_status` — Get the status of an async datasource-connection-test operation started via `test_datasource_connection`.
* `onboard_discovered_datasets` — Trigger async onboarding of one or more discovered datasets for a datasource. Onboarding promotes a discovered-but-not-yet-tracked dataset into a regular Soda dataset that can carry contracts, checks, and monitors.
* `get_onboard_discovered_datasets_status` — Get the status of an async dataset-onboarding operation started via `onboard_discovered_datasets`.

</details>

<details>

<summary>Discovered Datasets - tables found by discovery scans</summary>

Discovered datasets have been found by Soda but have not yet been onboarded as regular Soda datasets.

* `list_discovered_datasets` — List datasets Soda has discovered in your data sources. Discovered datasets have been found during a discovery scan but may not yet be onboarded as first-class Soda datasets.

</details>

<details>

<summary>Runners - execute scans and datasource operations</summary>

Self-hosted runners require API key credentials. Create them through the local-only `create_runner` tool, or follow the [Soda Runner deployment guide](broken://pages/Db0zxXDkPOUSKNNBuQXE#deploy-using-cli-only) to obtain them in Soda Cloud.

* `list_runners` — List Soda runners in your organization.
* `create_runner` — **Local only.** Create API key credentials for a new self-hosted Soda runner deployment.
* `get_runner` — Get a runner by ID, including online status, runner type, version information, and last-seen timestamp.
* `delete_runner` — Delete a self-hosted runner by ID.

</details>

<details>

<summary>Secrets - encrypted credentials referenced from datasource YAML</summary>

Datasource YAML can reference stored credentials as `${secret.NAME}`. The Soda Cloud Web UI is the safer place for a person to enter a secret because it keeps plaintext out of chat history and the MCP client/server flow.

`create_secret` and `update_secret` are available only through the local server. They accept plaintext locally and encrypt the value before transmitting it to Soda Cloud. Remote connections can list secret metadata and delete secrets, but cannot create or update their values.

* `list_secrets` — List secrets in your organization.
* `create_secret` — **Local only.** Create a new secret that can be referenced in datasource YAML as `${secret.NAME}`.
* `delete_secret` — Delete an existing secret.
* `update_secret` — **Local only.** Update the value of an existing secret. The secret name cannot be changed.

**Local secrets threat model**

* Plaintext exists in local agent process memory and on the local stdio pipe, but is never logged or sent to Soda Cloud unencrypted.
* The server uses AES-256-GCM and wraps the key with Soda Cloud's RSA public key. It caches that public key for about an hour and refreshes it once after an `encryption_failed` response.
* This is defense in depth over TLS. It does not protect against a compromised MCP client or local agent process.

</details>

<details>

<summary>Incidents - triage state on data quality failures</summary>

* `update_incident` — Update an incident's title, severity, status, description, resolution notes, lead, or linked check results.

</details>

<details>

<summary>Scans - observe and control scan executions</summary>

Contract, datasource, monitor, and data-standard tools trigger scans. These tools inspect or cancel those scan executions.

* `get_scan_status` — Get the current state of a scan. Use this to monitor scan progress during execution.
* `get_scan_logs` — Get log details for a scan in a final state. Use this to investigate issues with scan execution; log availability does not establish successful completion.
* `cancel_scan` — Cancel a running scan.

</details>

<details>

<summary>Users - organization users and groups</summary>

* `list_users` — List users in your Soda Cloud organization. Search by first name, last name, or email.
* `create_user` — Invite one or more users to your Soda Cloud organization.
* `disable_user` — Disable a user in your Soda Cloud organization.
* `list_user_groups` — List user groups in your Soda Cloud organization, including lists of members. Fuzzy search on group name.
* `get_user_group` — Get a single user group by ID, including its members. Returns the full group details (non-paginated).
* `create_user_group` — Create a new user group. Use this tool whenever the user asks to create a user group (typically naming a new group), even when the prompt also lists initial members.
* `update_user_group` — Update an existing user group's members. Use this tool only when the user references an existing group (by ID or by name); for a brand-new group, use `create_user_group` instead.
* `delete_user_group` — Delete a user group by ID. This action is permanent and cannot be undone.

</details>

<details>

<summary>Licensing - SPU consumption</summary>

SPUs (Soda Processing Units) are the consumption unit of SPU-based licensing. Consumption accumulates per day. An organization on any other licensing model is refused with a `400 organisation_not_on_spu_plan` rather than reading zero, so a zero total always means nothing was consumed over the period asked about.

The per-dataset breakdown requires **Manage organization settings** permission and accepts periods of at most 92 days. Read every cursor page, but count the repeated organization total and unattributed remainder only once. Keep unattributed consumption separate. To group by a dataset attribute, join all consumption rows to current `list_datasets` attributes, retaining deleted or unmatched rows separately. Multi-select groups overlap and are not additive; historical usage is grouped by today’s attributes.

* `get_spu_consumption` — Get the number of SPUs (Soda Processing Units) your organization consumed over a period.
* `get_spu_consumption_by_dataset` — Get the SPUs (Soda Processing Units) your organization consumed over a period, broken down per dataset.

</details>

<br>

***

{% hint style="info" %}
You are **not logged in to Soda** and are viewing the default public documentation.

Learn more about [Licensing & documentation access](https://docs.soda.io/soda-documentation/govern/security-licensing-and-compliance/documentation-access-and-licensing.md) or <a href="https://docs.soda.io/~gitbook/auth/login" class="button primary small">Log in to Soda Cloud</a>.
{% endhint %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.soda.io/soda-documentation/soda-ai/soda-mcp/mcp-tools.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
