Security
AI on ServiceNow with the control your security team expects
Connecting an assistant to your instances should not be a leap of faith. Here is how we protect credentials, what data we keep, what gets logged and which controls stay with you.
Key controls
Encrypted credentials
Each instance credential is encrypted at rest with Fernet (AES-128-CBC with HMAC), using a master key that does not live in the database. They are never shown in the panel or in responses.
OAuth 2.1 with PKCE
Claude connects through a standard OAuth flow with PKCE and dynamic client registration. Tokens expire and can be revoked at any time.
Exportable audit trail
Every execution is logged: tool, instance, user, date, result and a payload summary without credentials. Export it to CSV.
IP allowlist
The MCP endpoint can be restricted to authorized addresses or ranges, with an expiry date.
Production confirmation
Instances flagged as production ask for explicit confirmation before any write.
Least privilege
The technical user only has the roles you grant it in ServiceNow. We recommend starting on development or a PDI.
What data is stored
This is everything the service stores. Full details are in the privacy policy.
| Panel account | Email, password hash (bcrypt), role, creation date and last login. |
|---|---|
| Instances | Alias, URL, technical user and encrypted credential. |
| OAuth authorization | Registered clients, access and refresh tokens, scopes and source IP. |
| Audit | Tool, instance, user, date, result and a payload summary without credentials. |
| Working memory | Only the observations the user or the assistant explicitly save. |
| Cache | Table metadata and public ServiceNow documentation, with expiry. |
What is not stored
- The content of conversations with the assistant.
- ServiceNow records returned by a query, unless the user saves them as an observation.
- No data is used for advertising, profiling or model training.
ServiceNow permissions
The technical user needs the admin role or, at minimum, rest_api_explorer plus the x_mcp_bridge.executor and x_mcp_bridge.script_runner roles once the bridge is installed.
Infrastructure and operations
- HTTPS-only traffic.
- Panel with CSP, HSTS, X-Frame-Options and CSRF protection.
- PostgreSQL database under the control of OPTIMO TECH S.A.S.
- Instances unused for 30 days are invalidated and must be revalidated.
- Access, correction or deletion requests handled within 30 days.
Legal backing
- Commercial license from OPTIMO TECH S.A.S. (tax ID 901.957.428-4).
- Privacy policy under Colombia’s Law 1581 of 2012.
- NDA and data processing agreement (DPA) for enterprises.
IT security questions
Could the assistant delete production data by mistake?
Instances flagged as production ask for explicit confirmation before any write, and every operation is logged. The technical user also only has the roles you grant it.
Do you store records from my instance?
No. Records returned by a query go to the MCP client and are not kept, unless you explicitly save them as an observation in the client memory.
Can I restrict which networks can use it?
Yes. The MCP endpoint supports an allowlist of IP addresses or ranges, with an expiry date.
Do you use my data to train models?
No. Data is not used for advertising, profiling or model training. Processing by the assistant provider is governed by its own policy.
Does your security team have more questions?
Book a session with our technical team. We bring the datasheet and answer your vendor questionnaire.
Book a demo