Updated 2026-09-29
Security & data handling
What SmixAI reads, how it's protected, and how tenants stay isolated.
Overview
SmixAI is designed to give you insight while limiting what it reads to what a task needs, and it isolates every organization's data.
What SmixAI reads
- Discovery reads: object/field definitions, record counts, automation & code metadata, org limits (storage), license totals, and user activity (e.g. last login). Duplicate detection uses aggregate counts, not row exports.
- Agents read only what a task needs: SmixAI reads only the records it needs to do the task you ask for.
See What gets inventoried for the full list.
How your data is protected
| Control | Detail |
|---|---|
| Credential encryption | OAuth tokens, secrets, and private keys are encrypted at rest. |
| Tenant isolation | Each customer's data is kept separate from every other customer's, and we test that separation. |
| Least privilege | Connect with a scoped/integration user; server-to-server runs as a defined identity. |
| Audit trail | Significant actions are logged (see Activity & audit log). |
| Permissioned actions | Agents act only within the permissions you give them. New agents propose changes for your approval, and every production deployment needs a person's approval. |
How to: keep your connection secure
- Use a dedicated, least-privilege Salesforce user for scanning.
- Prefer JWT Bearer for automated scans so no interactive password is involved.
- Rotate credentials periodically (add a new credential, set primary, remove the old).
Use cases
- Security review: confirm what SmixAI reads for discovery, how agent actions are permissioned, and that credentials are encrypted.
- Procurement: point your security team here plus to your contract/DPA for specifics.
Notes
- For formal security documentation (certifications, DPA), contact the SmixAI team.