Security & governance

Security and governance for AI on your data

Every AI agent on Dataki works inside clear limits: your people's own access, read-only connections, a daily budget and a full audit trail. Here is exactly how, for whoever signs off.

01

Your credentials never leave Dataki

Database passwords and service-account keys are encrypted with Google Cloud KMS and decrypted only inside our backend, for the length of a single query. AI assistants connected to Dataki get answers, never connection strings, and no tool exists that could hand one over.

  • Warehouse credentials encrypted at rest with Cloud KMS
  • Google projects and Supabase databases connect by signing in, with no keys to paste; any database password you enter is encrypted with Cloud KMS too
  • No tool in the MCP server exposes or exports a credential

02

An agent inherits its user's access, and nothing more

Every API key acts as the person who created it. It reaches only the data sources that person can already open, and every request passes the same permission checks as a signed-in browser, at every step of the agent's own work.

  • Each key gets only the permissions you grant: listing sources, adding files, querying, reading dashboards, writing dashboards and running the agent are separate
  • A key can be restricted to one team even when its owner belongs to several
  • Creating keys, inviting users, linking cloud projects and storing OAuth credentials all need a signed-in person

03

Your people sign in with their company account

Teams on Microsoft Entra ID sign in through their own tenant. The password, the MFA prompt and your Conditional Access policies all happen on Microsoft's side, against your directory, and you decide there who may use Dataki. Each person lands in your team, identified by the IDs in a token your tenant signed, never by an email address alone.

  • Set up per team: you register Dataki in your own Entra directory and we connect it to your team; its client secret is encrypted with Cloud KMS
  • Returning people are matched on their tenant and object IDs, which never change
  • An email address counts only on the domains your team lists, so a tenant can speak only for its own people
  • Each sign-in can be completed once, only in the browser tab that started it, and within ten minutes
Dataki's sign-in page with Sign in with SSO open: a Work email field, a Continue button, and a link back to other sign-in options
Enter a work email and Dataki sends you straight to your company's sign-in page.

04

Every query is recorded

Each query is logged with who or what ran it, which key, the SQL, how many rows came back and how long it took. Failures are logged too, so an agent that misbehaves is easy to spot.

  • Per-query attribution down to the individual API key
  • Each person reads their own log in the app, under API keys, or pulls it through the API
  • Kept after a key is revoked, so its full history stays on record

05

Every agent runs on a daily budget

Every API key has a daily limit on queries run and on data scanned, so an agent's spend is capped before it starts. When a key reaches its limit, further queries are refused with a clear message the assistant passes on to its user, and the limit resets at midnight UTC.

  • Daily caps on query count and bytes scanned, per key
  • Enforced on every query, as it runs
  • Set per key, so a shared production key and a personal experiment each get the right limit

06

Revocation is immediate

Revoke a key and it stops working on the very next request. Keys are checked on every request and never cached. The key's record stays, so its audit history still shows whose key it was.

  • Takes effect on the next request, with no propagation delay
  • Keys can be given an expiry date at creation
  • We store only a one-way hash of each key, so even a copy of our database holds no usable credentials

Want the details?

The full key model is documented: scopes, storage, error codes and what each route accepts. Ask and we will send it, or answer your question directly.