Automating Data 360

Three routes

Route	Use it to	Covers
Connect API (/services/data/v66.0/ssot/...)	Build and run inside one org, from a script or CI	Connections, data streams and their runs, DLOs, DMOs, mappings, data spaces, identity resolution and runs, calculated insights and runs, search indexes, data graphs and refreshes, segments, query-sql, data kits
Data kits	Move a working configuration to another org	Stream bundles (CRM and others), DLOs, transforms, calculated insights, identity resolution, search (DataSemanticSearch), AI retrievers, segments, data actions, semantic models and more
Metadata API (sf project deploy)	Version and deploy what a kit packages	DataPackageKitDefinition, DataPackageKitObject, DataKitObjectTemplate, DataStreamDefinition, DataSrcDataModelFieldMap, ObjectSourceTargetMap, MktCalcInsightObjectDef, DataObjectSearchIndexConf, MarketSegmentDefinition

There is no metadata type for an identity resolution ruleset or a retriever in Salesforce's registry. Those move through a data kit, or through the Connect API with a captured body.

Data kits

1. In the source org, add the components to a data kit. A kit is metadata (DataPackageKitDefinition): retrieve it, commit it and deploy it.
2. In the target org, deploy the kit's components with POST /services/data/v66.0/ssot/data-kits/{dataKitDevName} and a components array of { "type", "config" } objects. Configs from Salesforce's payload guide:

type	config fields
DataStreamBundle (CRM)	bundleName, forceNoRefresh, bundleConfig.orgId
DataLakeObject	dataSourceObjectDevName, apiName, label, dataSpaceName
IdentityResolution	dataSpaceName, templateDevName, dataKitDevName
CalculatedInsight	apiName, apiNameOverride, label, publishInterval
DataTransform (batch)	dataTransformDevName, apiName, label, dataSpaceName, overrides
DataSemanticSearch	dataSpaceName, searchIndexName, dataKitName
MarketSegment	name, dataKitName

AI Retriever, Activation, Data Custom Code and others are supported too, but the guide gives no example payload for them.

Related endpoints:

• GET /ssot/data-kits and /ssot/data-kits/available-components;
• GET /ssot/datakit/{name}/manifest (note the singular datakit);
• POST …/data-kits/{name}/undeploy;
• per-component …/components/{component}/dependencies and …/deployment-status.

Per a community audit of Salesforce's v67 spec, there is no get-single-kit or deploy-job status resource: poll component status instead. That audit also found that deploying a kit into the org that built it was rejected. Deploy to a different org.

Capture and replay

The nested request bodies for identity resolution, search indexes and data graphs didn't render for the hub in Salesforce's Connect API reference. The reference page itself (developer.salesforce.com/docs/data/connectapi/references/spec) has the OpenAPI document, about 201 operations at v67 per the community audit; download it there. For a script that must not guess:

1. Build the component once in a development org, in the UI or by API.
2. GET it and remove server-owned fields: IDs, status, run times, created and modified stamps.
3. Commit the JSON next to the pipeline.
4. In other orgs, look it up by name, and POST it if it's absent or PATCH it if it differs.

guides/data360/automate-the-golden-record-pipeline.md implements this.

Order of operations

1. Manual and Setup:

• Data 360 provisioned;
• Einstein and Agentforce on (guides/agentforce/build-agent-with-agentforce-dx.md);
• the integration user's Data 360 permission sets;
• the CRM connection;
• the agent user's Data 360 access and data space scope.

1. Streams for Account, Contact, Order, OrderItem and Product2 → wait for the first refresh.
2. Mappings, including Order → Sales Order and OrderItem → Sales Order Product.
3. Identity resolution rulesets → run → wait (up to 30 minutes).
4. Calculated insights → run.
5. Narrative DMO and its transform → run. Then the search index → Ready, the data graph → refresh, and the retriever → active.
6. Data Library (sf agent adl create) → agent publish and activate → agent tests.

Rules for the scripts

• Idempotent: every step reads before it writes, and the script can be re-run after any failure.
• Gate the credit-consuming steps behind an explicit flag: stream runs, identity resolution runs, insight runs, index builds and data graph refreshes all consume Data 360 credits (family 10).
• Respect run limits: identity resolution on-demand runs are limited per day (reference/data360/identity-resolution.md).
• Verify with SQL after each stage: counts in the DLO, DMO, unified, CIO and chunk objects, through query-sql.
• Tools: no official sf plugin manages Data 360 setup. A community plugin, @anushdsouza/sf-plugin-data360, wraps 135 operations and is an experimental beta, not a Salesforce product. Its source is a useful map of endpoint paths.