Written to be forwarded to your security team.
What the model sees, who can do what, what is recorded and where it runs, stated plainly.
Personal data is masked before anything reaches a model.
The platform connects to your database with read-only access. To answer a question, the model sees your table and column names, your question and a capped slice of the result, with personal data masked before it is sent. Your team sees the real values; the model doesn’t.
Secret-grade data, such as card numbers and national IDs, stays masked for everyone. When the agents build a pipeline, the model works from the shape of your data, and personal values are withheld. Questions are screened before any AI step runs, and answers are checked on the way out.
Plainly: The answers your team sees, real names included, are kept in the workspace’s conversation history, on our side.
“Which customers ordered most last quarter, by region?”
| full_name | region | orders | |
|---|---|---|---|
| masked | masked | Northwest | 214 |
| masked | masked | Southwest | 198 |
| masked | masked | Midwest | 187 |
Access is by invitation, and every role is spelled out.
There is no self-sign-up. An admin adds each person, who sets their own password on first sign-in, and the roles are built in, the same on every plan. With teams on, members see only their own team’s work.
The Auditor reads everything, AI FinOps included, and can change nothing.
Plainly: Single sign-on and multi-factor sign-in are available on request.
- Owner
- Runs the organization. The only role that can grant Owner or Auditor.
- Admin
- Invites people, sees every project and AI FinOps.
- Auditor
- Reads everything. Changes nothing.
- Builder
- Builds and runs pipelines for a team.
- Approver
- Approves spend on builds and changes, when spend approval is on.
- Viewer
- Reads a team’s work.
Every answer and every change is on the record.
Each answer is recorded: who asked what, where, whether personal data was involved, and enough to rebuild the answer end to end.
Every pipeline change is a pull request a person on your team approves, and deploying is a separate click. Each pipeline keeps a dated history, from the request in your words to every release.
AI FinOps shows what each pipeline and workspace has used. Retries on our side are ours, and never count.
Plainly: None of these records is tamper-proof today, and we don’t describe them that way.
- Revenue by region, last quartera.khan · sales workspaceanswered
- Top 10 customers by spendm.ortiz · sales workspace · personal data masked for the modelanswered
- Pull request approved and mergedj.lee · your repositorymerged
- Deployedj.lee · a separate clickdeployed
We host the platform. Your code and pipelines stay with you.
Every build lands as code in your repository and runs in your cloud.
We support the databases, warehouses and clouds you already run: PostgreSQL, SQL Server, Oracle and MySQL; BigQuery, Snowflake, Databricks and Redshift; Google Cloud, AWS and Azure.
- Your code
- Readable code in your repository. Every change is a pull request your team approves.
- Live
- Your pipelines
- Run in your cloud, on your schedule.
- Live
- Managed service
- Hosted by us, in the US. Database credentials kept in a secrets vault.
- Live
- Dedicated deployment
- Your organization on its own deployment, with AI included or your own model key.
- On request
- Air-gapped
- Talk to us about your requirements.
- On request
- Single sign-on and MFA
- Set up with your identity provider.
- On request
- Certifications
- SOC 2.
- In progress
Found a security issue? Tell us first.
Email us with the details. We’ll acknowledge your report and work with you on a fix. Please give us a reasonable window before any public disclosure.
- To
- security@nectorq.ai
- Subject
- Security report: one-line summary
- What you found
- Where you found it
- Steps to reproduce
- What someone could do with it
Send this page to your security team, then bring us their questions.
Book a demo and we’ll walk them through these controls in the product. A detailed security overview is available under NDA.