Your data stays in your environment.

Opswell builds automations inside your Claude workspace and your systems. We do not move client production data into Opswell-owned multi-tenant vaults.

Principles

Your tenant. Work runs in your Anthropic / Claude org and your apps.
Least privilege. Named users or service accounts; access revoked at handover.
Human in the loop. Nothing posts to GL / ERP / CRM unreviewed unless you approve an exception in writing.
Audit trail. Preparer / reviewer / approver separated where controls require it.

Claude & licensing

Builds are Claude-native (Anthropic). We are independent — not affiliated with Anthropic.
Client work should run on Team / Enterprise (or equivalent) with data-use settings that match your policy.
You carry the Claude license; model usage is billed by Anthropic to your org.

Regulated data

HIPAA / BAA: Healthcare PHI needs a BAA first. CUI stays out of cloud AI unless your security officer approves in writing.
CUI / GovCon: Discussed up front; we will not invent a compliance posture we cannot support.
SOC 2: Not held today — work runs in your tenant under NDA/MSA with least-privilege access.
Typical fit: AI automation for companies roughly $10M–$500M revenue (and PE-backed portcos in that band); finance leadership from about $2M / Series A onward. Security depth scales to the engagement.

Access, NDAs, offboarding

NDA on request before any system access.
Access limited to named people for the engagement window.
At close: credentials revoked, artifacts delivered to you, Opswell retains no production copies.