Salesforce MCP Server

Salesforce

Salesforce is a CRM where revenue teams track accounts, contacts, leads, and opportunities through the sales pipeline. Custom objects extend the same record model in each org.

First moves

1. Accounts — list or open company records when the user names a customer or territory.
2. Contacts — list or open people tied to accounts for outreach and account planning.
3. Leads — list or open pre-qualified prospects before conversion into accounts.
4. Opportunities — list or open pipeline deals with stage, amount, and close date.
5. Object catalog — discover which record types exist in the org before custom-object work.
6. SOQL reads — ad-hoc SELECT for joins, filters, and rollups (COUNT, SUM, GROUP BY) when list views are too narrow.
7. Keyword search — cross-object search when the user names a term, not a specific table.
8. Schema inspection — field names, picklists, and relationships for an object before writing filters or updates.

Data model

Account → Contact / Opportunity. Accounts represent companies and can sit in parent/child hierarchies. Contacts are people, usually linked to an account. Leads are prospects before conversion. Opportunities are active deals moving through pipeline stages.

Operational order

1. Start from the object the user named — account planning begins with the account, pipeline reviews with opportunities.
2. Use SOQL when you need aggregates, date windows, or joins across related objects.
3. Use keyword search to locate ids when the user only has a name or phrase.
4. Inspect schema on unfamiliar or custom objects before building filters or writes.
5. Continue paginated SOQL from the prior result set instead of re-running the same query when more rows remain.

Writes

Create, update, delete, and upsert records by object type and id. SOQL and keyword search are read-only — discover record ids first, then write. Destructive deletes need explicit user approval.

Gotchas

• Governor limits — SOQL returns at most 2000 rows per query; paginate or narrow filters for large orgs.
• Aggregate SOQL — COUNT, SUM, and GROUP BY are supported for pipeline rollups but remain read-only.
• Default field sets — standard list reads return a small column set; name the fields you need in SOQL for wider projections.
• Custom objects — every org differs; inspect schema before assuming field or picklist names.
• Schema weight — object metadata responses are large; reuse field names within a session instead of re-inspecting every turn.
• Lead vs contact — leads are pre-conversion; converted leads become accounts and contacts with new ids.