Authenticate via OAuth 2.0 to obtain a Salesforce access token using the connected app client credentials flow
POST to /services/data/vXX.0/sobjects/Case with required fields (Subject, Description, Origin, AccountId) to create the case
Verify Einstein Case Classification is enabled in Setup and that a prediction model has been trained on sufficient closed cases
Wait for asynchronous Einstein classification — the model auto-populates picklist/checkbox fields (e.g. Type, Reason, Priority) and triggers assignment rules
GET /services/data/vXX.0/sobjects/Case/{caseId} to confirm predicted field values were written back to the case
Update any field Einstein left below threshold confidence using PATCH /services/data/vXX.0/sobjects/Case/{caseId}
Known gotchas
Einstein classification runs asynchronously after case save — polling immediately after POST will likely return blank classification fields; build in a short delay or use a platform event trigger
The prediction model requires a minimum number of closed cases with consistent field values to produce reliable suggestions; low data volume returns no predictions silently
Field predictions only work on picklist, checkbox, and lookup fields — free-text fields like Subject are not predictable and will not be populated
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp