datatape

No arbitrary SQL. Ever.

Datatape enforces security at the platform level — not by convention, by architecture. AI agents can only call tools you define, with SQL templates you wrote, using parameters you validated. They never touch your credentials or construct their own queries.

How Datatape handles your data

Database credentials
Encrypted at rest with Fernet (AES-128-CBC + HMAC). Never returned in API responses. Never logged. Never accessible to AI agents or end users.
Query execution
Datatape connects to your database, executes the pre-defined SQL template, and returns results. Data flows through — it's not stored permanently.
Query results
Returned to the caller via MCP or API. Not cached by default. Not used for training. Not shared across organizations.
SQL safety
Validated by SQLGlot to be read-only. INSERT, UPDATE, DELETE, DROP are rejected at the platform level. Enforced server-side — cannot be bypassed by AI agents.
Parameters
Parameterized queries prevent SQL injection. User inputs are never interpolated directly. Jinja2 templates rendered server-side with escaped values.

Access control

Role-based access (RBAC)

Admin — manages connections, tools, servers, API keys, and team members
Editor — creates and modifies tools and servers
Viewer — can view tools and servers but cannot modify

Authentication

WorkOS-powered SSO (SAML, OIDC, Google, Microsoft, GitHub)

API keys with dt_ prefix, scoped per server

MCP endpoint auth via Bearer token or API key

API key isolation

Each API key is scoped to a specific server

Keys are hashed (SHA-256) before storage — raw keys cannot be retrieved after creation

Keys can be revoked instantly

Encryption

Layer
Method
In transit
TLS 1.2+ for all connections (API, MCP, dashboard)
Credentials at rest
Fernet symmetric encryption (AES-128-CBC + HMAC-SHA256)
API keys
SHA-256 hashed before storage

Audit logging

Every tool execution is recorded

What's logged

  • Who — authenticated user or API key
  • What — which tool, with what parameters
  • When — timestamp
  • Result — success/failure, error message, row count, execution time
  • Where — which MCP server, which database connection

Retention by plan

Team7 days
Scale90 days
Enterprise1 year (configurable)

Infrastructure

No shared database connections between organizations
No customer credentials in logs, error messages, or stack traces

Shared responsibility

Datatape is responsible for
You are responsible for
Encrypting your database credentials at rest
Granting read-only database credentials
Enforcing read-only SQL validation and template-only execution
Defining which SQL queries are exposed as tools
Authenticating users and API keys
Managing who on your team has access
Logging every tool execution
Reviewing audit logs for your organization
Securing the Datatape platform and infrastructure
Securing your database (network rules, firewall, user permissions)
Patching platform vulnerabilities
Rotating database credentials periodically

Security FAQ

No. Query results are returned to the caller and not persisted by Datatape. They are not cached, not used for training, and not shared across organizations.

No. AI agents can only call tools you define. Each tool has a fixed SQL template with parameterized inputs. The SQL is validated by SQLGlot to be read-only — INSERT, UPDATE, DELETE, and DROP are rejected at the platform level.

Datatape employees cannot see your query results. Database credentials are encrypted and not accessible in plaintext through any admin interface. Infrastructure access is limited to essential personnel and logged.

Tools stop working immediately. Datatape has no cached copy of your data. Revoke the database credentials you provided, and access is fully terminated.

Yes. Datatape uses WorkOS for authentication, which supports SAML 2.0, OIDC, Google Workspace, Microsoft Entra ID, and GitHub. SSO is available on the Enterprise plan.

Email security@datatape.io. We take all reports seriously and will respond within 48 hours. We welcome responsible disclosure and will not take legal action against good-faith security research.

Security inquiries

For security questions, vulnerability reports, or to request a pre-filled security questionnaire.

security@datatape.io
Vulnerability reports: 48h responseQuestionnaires: 5 business days