Skip to content
SayDov
NL Nederlands
SayDov

Security

Security and privacy

For IT and privacy officers: this is how we set up Claude on your Salesforce, and how you stay in control.

Last updated:

How the setup works

Every colleague who works with SayDov uses Claude Code: Claude’s working environment on their own computer, or the Code tab in the Claude desktop app. Claude works with Salesforce through the official Salesforce CLI, signed in with that colleague’s own account. All rules, ways of working and changes live in a GitHub repository in your own organisation.

Sandbox first

  • By default, Claude only works in a sandbox: a test copy of your Salesforce.
  • Claude gets no access to production. A change only goes to production after someone on your team has reviewed and approved it, using your usual process.
  • New automations (flows) are saved as drafts. A person switches them on.

Minimal permissions

  • Every user signs in with their own Salesforce account. No passwords are shared.
  • That account only gets the permissions it needs, through dedicated permission sets.
  • Signing in uses Salesforce’s official login flow (OAuth). Claude never sees passwords.

A safety filter

On top of the agreements in the rulebook, an automatic filter checks every command before it runs. Among other things, it blocks:

  • signing in to production, or changing anything there;
  • changing or deleting data;
  • changes to profiles, sharing rules and security settings;
  • deleting metadata.

We tailor what is and isn’t allowed to your organisation. The rules and the filter live in your own repository. Claude can’t change them unnoticed: every change asks for your approval. After handover, you can adjust the rules yourselves.

Full history and rollback

  • Every change is recorded in Git: who, what, when and why.
  • You can always roll back to a previous version.
  • Your team can review everything before it goes to production.

Your data at Anthropic

We work with Claude Team or Claude Enterprise. Anthropic’s commercial terms apply to both:

  • Anthropic does not use your data to train models, unless you choose to allow it.
  • A data processing agreement (DPA) with the European Commission’s standard contractual clauses is part of those terms.
  • Data is kept for 30 days by default. With Claude Enterprise, ‘zero data retention’ is possible in certain cases.
  • Anthropic holds ISO 27001, ISO 42001 and SOC 2 Type II certifications, among others. See Anthropic’s Trust Center.

We never use consumer plans (Free, Pro and Max) for company data.

Processing in the EU

The standard Claude services process data globally or in the United States. Do you want everything to stay in Europe? Then Claude Code can run through AWS Bedrock or Google Cloud Vertex AI, in a European region. We discuss this during the intro call.

Data on the user’s computer

Claude Code keeps conversations locally on the user’s computer for a while. We set a short retention period and switch off optional telemetry.

SayDov and your data

  • We sign a data processing agreement with every client.
  • We work in your sandbox, with the accounts you give us.
  • After the 6 months of support, you revoke our access: to GitHub, Salesforce and any other systems.

As of October 2026. The terms of Anthropic, Salesforce and GitHub may change; we check them again at the start of every project. Questions about security? Email us at info@saydov.com.