Trust

Who signs in

I ask you to sign in with your company Microsoft account. By default, I refuse personal accounts.

Each person connects with their own Oracle account, so Oracle applies their own roles. If an admin shares a connection, everyone who uses it acts with the roles of the account on that connection.

What it can change

I check every Oracle call against your organization's policy. New organizations start with writes off, and admins control them on the Policies page.

I return a preview before every allowed write. The write needs that preview's single-use confirmation token within ten minutes. Environment administration never runs on production.

What it keeps

I encrypt Oracle credentials with a key the database does not hold. I keep an activity log of who called which tool, when and the outcome, never cell values.

I keep a fingerprint of each result for 14 days by default, so a repeat read can say what changed without keeping the numbers. I keep the numbers only when your admin turns on result reuse. I also keep your policies, saved definitions, feeds and schedules.

What FMEPM sees

To fix problems, FMEPM sees which tool failed, the error code and when, never your numbers, member names or Oracle's error text.

On the demo environment I keep the full error text, because it is sample data.

Where it runs

Trials run on the hosted service, which keeps its data in Canada. For customers, FMEPM can run it in your own cloud, in the region you choose.

Your conversation with the AI app is processed by the AI company under your agreement with them; I never send it Oracle credentials.

The AI

I never send Oracle credentials to the AI app. I treat Oracle text, including member names, aliases, job messages and errors, as data, never as instructions.

I call no AI model myself; my rules are fixed code.

Limits

My default limits are 60 calls per minute, 5,000 calls per day and 500,000 cells per export. Your plan can set different limits, and admins can lower them.

Certifications

FMEPM Connect is not SOC 2 or ISO 27001 certified today. Full security details are for customers, or email security@fmepm.com.

Reporting a security problem

Email security@fmepm.com with the steps to reproduce. You get a first response within two business days, and a fix date or a reason once I have looked. Please do not test against another customer's organization; the demo environment is there for that.

Questions

Can the AI see my Oracle password?

No. I encrypt your Oracle credentials and never send them to the AI app.

Can a user do more through Claude than in Oracle?

No. Oracle applies the roles of the Oracle account on the connection. With your own connection that is you; with a connection an admin shared, it is the account on that connection.

Can I turn writes off?

Yes. Your organization admin can turn writes off on the Policies page, and a stricter profile can block them too.

Does FMEPM see my numbers?

My operator error page hides your numbers, member names and Oracle error text for test and production connections. I keep numbers with remembered results only when your admin enables result caching.

What happens when someone asks for a change on production?

I check your organization's policy first. If it allows the write, I return a preview and require its single-use confirmation token within ten minutes. Environment administration is always refused on production.