Control Access
How Credible secures and governs data - one enforcement point for every query, sign-in options, API keys, roles, access rules in the data model, data protection, audit logs and plan availability
Credible checks every query against the caller's identity, their roles and the access rules in the data model before it reaches your database.
How Access Is Enforced
Every query reaches your data through Credible, whichever surface sends it: workspace chat, Slack, an agent over MCP, a dashboard, a data app or the REST APIs. On each request Credible:
- Identifies the caller: a signed-in person, or an API key acting as a group.
- Checks the caller's role on the organization, environment, package or workspace.
- Applies the row and column access rules in the data model, with secure givens filled from the caller's verified identity, so the caller can't supply them.
- Runs the query with the environment's stored connection, and records the request.
Because there is one enforcement point, there is no separate set of permissions per tool to keep in sync. The rules apply to every query that goes through a data model. Two kinds of access don't:
- Raw SQL on a connection. Environment Admins and Modelers can run SQL directly on the environment's connections, for example to explore tables while modeling. That SQL doesn't go through a model, so access rules don't apply to it; it reads whatever the connection's credentials can read. Give these roles only to people who may see all of that data, or connect with credentials scoped to what they may see.
- Direct warehouse access. A BI tool or SQL client connected straight to your warehouse is governed by the warehouse's own permissions.
Governed Definitions
- Definitions live in the data model. Metrics, joins and access rules are defined in the data model, the one place every surface reads them from. The model is stored as code and can be reviewed like code, for example in pull requests.
- Every change is a versioned package. Publishing creates a new, immutable package version. One version is served as latest, and you can switch back to an earlier one. See Publishing and Versions.
- Access rules ship with the model. Row and column access rules are versioned, published and rolled back together with the definitions they protect.
Identity and Sign-In
People sign in with one of:
- Microsoft account
- GitHub
Credible configures sign-in methods; there are no sign-in settings in the app.
Signing in proves who someone is; it doesn't grant access. An organization admin adds each person to the organization with an organization role in Users and Groups, and access to environments, packages and workspaces is granted as described in Permissions.
Enterprise SSOEnterprise
People can also sign in through your company's identity provider:
- Microsoft Entra ID. An administrator of your Entra ID tenant approves the Credible app, and then people sign in with their work accounts. See Single Sign-On with Entra ID.
- Okta. Credible sets up a sign-in connection for your organization.
- SAML. For other identity providers, on request: ask Credible whether yours is supported.
Contact Credible to set up Okta or to ask about SAML.
API Keys
Services, scripts, CI/CD jobs and embedded products authenticate with group API keys (group access tokens), sent as Authorization: ApiKey <key>. Organization admins create them on a group's page in the app or with cred add group-access-token; group admins and organization admins can revoke them. The secret is shown only when the key is created.
A key acts as its group, with every role the group holds, including organization Admin. Keys have no scopes of their own, and there are no personal keys. To limit what a key can do, give it a dedicated group with only the grants the integration needs. See What a Key Can Reach.
MCP clients authenticate with OAuth as the signed-in person, or with a group API key. Workspace and package MCP URLs narrow what an agent sees, but they are not an access boundary: the caller's roles are. See REST APIs.
Roles and Access Rules
Access has two layers:
| Layer | Decides | Where it is set | Page |
|---|---|---|---|
| Resource roles | Who can administer, model in, query or share an organization, environment, package or workspace | The app, the CLI or the Admin API | Permissions |
| Access rules in the model | Which sources a caller may query and which rows they see | #(authorize) and #(access_filter) annotations on sources | Row and Column Access |
Access rules read the caller's identity from secure givens: the caller's groups, or values an organization admin assigns per user, group or everyone on the Access Control page. For API keys only the groups apply, and that is how a product that embeds Credible keeps each of its customers to their own rows; see Tenant Isolation for Embedded Products.
Data Protection
- Database credentials stay in Credible. Connection credentials are stored in the environment. Only environment admins can read or change a connection's configuration; modelers can use a connection without seeing its credentials, and analysts never connect to the database. Credible runs every query on the caller's behalf.
- Encryption. Data is encrypted in transit with TLS 1.2 or higher and at rest with AES-256, as stated on the security page.
- Network access to your database. Queries leave from fixed egress addresses you can allowlist, and PostgreSQL can also be reached through an SSH bastion. PrivateLink, VPC connectivity and other private networking are part of the Enterprise plan, set up on request. See Connection Settings.
- Isolation. Organizations are separated by identity on shared infrastructure. The Enterprise plan adds dedicated single-tenant clusters.
- Materialized tables. Tables Credible builds for performance live in your warehouse or in Credible-managed storage. A built table behind an access rule holds every caller's rows, and the rule is applied each time it is read; see Value Search and Materialization.
- Compliance. Credible is SOC 2 compliant, covering the Security trust services criteria. Request the security package, including the auditor's report, from security@credibledata.com. See the security page, privacy policy and subprocessor list.
Audit LogsEnterprise
Credible records each request with the caller's identity, including which packages each caller read. There is no audit page in the app; an organization admin can ask for access to their organization's logs. To find or export your logs, contact support. See also Monitoring.
Plan Availability
Row and column access rules, audit logging, enterprise SSO (Entra ID, Okta and SAML), private networking and dedicated clusters are part of the Enterprise plan. See Pricing.