Your Salesforce data holds the answers. Now you can ask.

How It Works

From Salesforce data to trusted answers in seconds

1. 1

   Connect

   Point Nexatron at your Salesforce org. Native SOQL and Bulk API connection, no middleware, no exports. Read-only. Respects your existing security.

2. 2

   Ask

   Ask any business question in plain English, in Nexatron or from Claude through MCP. Cross-object, cross-system, cross-period.

3. 3

   Trust

   Every answer comes with the SQL generated, sources queried, confidence score and a full audit trail. Governed by your semantic layer.

Which Salesforce controls still apply when Claude asks a question?

All of them. The MCP server is one more client of your org, and Salesforce treats it exactly like any other integration.

• OAuth scopes on the app

  The MCP server authenticates through an External Client App or a connected app, and the OAuth scopes you assign define which protected resources it can reach. The api scope covers REST API and Bulk API 2.0 access on behalf of the integration user. Scopes narrow what the app can touch; they never add permissions the integration user does not already have.

• Permission sets on the integration user

  Give the server its own integration user and grant object access through permission sets, the method Salesforce recommends. Salesforce also ships integration permission sets that define the scope of data access for integration services. Start with Read on the objects a pilot needs and nothing else.

• Field-level security

  Field permissions decide whether a user can view or edit each field, and Salesforce enforces them in its APIs as well as in the UI. A field hidden from the integration user never reaches Claude, whatever the prompt asks for.

• Sharing rules and org-wide defaults

  Org-wide defaults set the baseline for records a user does not own, and sharing rules can only grant additional access, never take it away. Claude sees the records the integration user is allowed to see and no others.

How should a Salesforce MCP server be run?

Read-only first, logged on both sides, and governed by one set of definitions.

• Read-only first

  Start with query tools only. Actions that write to Salesforce, such as updating an Opportunity stage or creating a Case, come later, one at a time, each with its own approval path and its own permission set.

• Log every request

  On the server side, record who asked, which query ran, which objects and fields it touched and what came back. On the Salesforce side, Event Monitoring (a Salesforce Shield or Event Monitoring add-on) shows who accessed which data, when and from where.

• Govern the definitions

  Permissions decide who can see a field. They do not decide what 'pipeline' or 'win rate' means. Nexatron's semantic layer holds one certified definition per metric, so Claude, a dashboard and an export all agree.

Sources: Salesforce, Salesforce DX MCP Server README on GitHub and Manage MCP Servers in API Catalog in Salesforce Help, and the Model Context Protocol introduction. Current Nexatron connector status is published at nexatron.io/connectors.

Use Cases

Questions your Salesforce team asks every week

• Question

  “What's our pipeline health by stage and region?”

  What Nexatron Returns

  Queries Opportunities across regions with stage-level breakdowns, win rates, and velocity metrics.

• Question

  “Which accounts have the highest expansion potential?”

  What Nexatron Returns

  Analyzes account history, usage patterns, and whitespace against similar accounts.

• Question

  “Show me customer churn risk by segment”

  What Nexatron Returns

  Combines Service Cloud cases, engagement scores, and contract renewal data.

• Question

  “What's our marketing-to-revenue attribution?”

  What Nexatron Returns

  Traces Campaign and Campaign Influence records through to closed-won Opportunities, using the attribution model configured in your org.

Capabilities

Built for Salesforce. Built for the enterprise.

• Native Salesforce Connector

  Direct SOQL and Bulk API connection. No middleware, no CSV exports, no workarounds.

• MCP Tools for Claude

  Governed query tools that Claude, ChatGPT and Copilot can call, with your definitions and permissions applied to every request.

• Cross-Object and Cross-System Analytics

  Query Sales Cloud and Service Cloud objects such as Opportunities, Accounts, Cases and Campaigns in a single request, and combine Salesforce with supported sources such as NetSuite, Oracle Fusion and Snowflake. Salesforce Marketing Cloud is Coming soon as a separate connector.

• Governed Semantic Layer

  Define 'pipeline', 'ARR' and 'churn rate' once. The definition is enforced across every query and every user.

• Field-Level and Row-Level Security

  Respects Salesforce org-level, field-level and record-sharing controls, and adds row-level policies of its own.

• Audit Trail

  Every answer includes the SQL generated, sources queried, confidence score and timestamps.

Resources

Go deeper on Salesforce, Claude and Agentforce

• GuideSalesforce MCP Server: Governed Access for ClaudeWhat a governed Salesforce MCP server has to do before Claude gets its first question: identity, scopes, logging and definitions.
• GuideAgentforce Implementation ChecklistThe decisions, permissions and tests to settle before an Agentforce agent goes live in Sales, Service or Marketing.
• GuideClaude Connectors for Enterprise SystemsHow Claude connectors reach ERP and CRM systems, and what to approve before a team enables one.
• ChecklistClaude connector approval checklistA printable checklist for approving a Claude connector: owner, identity, scopes, data classes, logging and review date.
• AgentsAgentforce implementation and Salesforce AI agentsAgentforce implementation and consulting, plus the Revenue Intelligence and Customer Health agents dotSolved.ai builds on top.
• HubClaude for the enterpriseThe Claude hub: deployment options, governed MCP access to NetSuite, Salesforce and Oracle Fusion, and the full Claude reading list.