Privacy Policy
Last updated 8 October 2026
What we collect
Account data: name, work email, hashed password, workspace name and your role. Billing data: plan, subscription status and a customer reference at our payment processor; card details never reach our servers. Connection data: database host, port, database name and username, plus the password encrypted with AES-256-GCM. Catalog data: table, column, key, index and routine metadata, comments, row counts and a bounded number of sample values used for type inference. Audit data: which tool ran, with which typed arguments, by whom, and the sanitized outcome. Audit records never contain credentials, connection strings or result rows.
Deployed Co-Pilots and your end users
When you embed a Co-Pilot in your own application, the questions its users ask, the answers given, the outcome and the tool used are stored in your workspace as conversations, together with an optional end-user identifier you choose to pass. Requests carry the browser’s origin for allow-list checks and are rate limited per key. We do not set cookies in your application; the widget keeps the current conversation id in the visitor’s session storage only. You are the controller of this data and responsible for informing your users.
Contact form
Messages sent through the contact page (name, email, optional company, topic and message) are emailed to our support mailbox and used only to reply to you. They are not added to marketing lists.
Background processing
Semantic enrichment and tool enhancement run as background jobs on our servers so that a closed browser tab does not waste your model credits. Their progress and results are stored in your workspace until you accept or discard them; nothing is sent to the model that is not listed in the next section.
What we do not collect
We do not copy your tables. Query results are streamed to your browser and capped at 100 rows; they are not stored by the Service. Your LLM API key is encrypted per workspace and is never logged.
How data is used
To provide the Service: discovery, semantic modelling, governed tool generation, drift review, Co-Pilot answers and billing. To secure it: authentication, rate limiting and abuse prevention. To communicate: verification codes, billing notices and the announcements your platform owner configures. We do not sell personal data and do not use your data to train models.
Language-model providers
When you configure an LLM key, we send only what the governed workflow needs: for routing, approved tool names, descriptions and parameter names; for argument extraction, the selected tool’s parameter contract; for answer wording, at most 20 aggregate rows; for semantic enrichment and tool enhancement, table, column and metric names and descriptions. Never your schema, SQL, credentials or row-level data. Your provider processes these requests under its own terms.
Retention
Account and workspace data are kept while your workspace exists and deleted within 30 days of termination. Tool execution audits are kept for the retention window of your plan and pruned automatically. Catalog snapshots are capped per data source.
Security
Transport is encrypted (TLS). Secrets are encrypted at rest with keys bound to the owning workspace. Database connections are opened read-only. Access to production systems is restricted and logged.
Your choices
You can update your profile and password, remove data sources and their credentials, rotate or delete your LLM key, cancel billing and close your workspace from the application. For access, correction or deletion requests, or any question about this policy, contact support.