Security model
Target Audience: Developers, security reviewers, and technical decision-makers integrating Toucan AI.
TL;DR
Toucan AI protects data with layered controls: short-lived embed tokens, organization isolation, row-level security on queries, encrypted database credentials, and clear boundaries between platform data and your connected data.
What this section covers
- Data access boundaries — who can access what
- Data storage & retention — what Toucan stores, query data flow, caching
- PII & personal data — where personal data can appear and how to minimize it
- Secrets & encryption — credentials and tokens
- Token security — embed authentication
- API security — access controls for APIs
- AI assistant data handling — conversations and AI workflows
- Third-party subprocessors — external providers that may process data
For embed-specific responsibilities, see also Security boundaries.
Data access boundaries
Token security
API security
MCP access and security
How the Maker MCP and the Explorer MCP authenticate, what each credential can do, how long it lasts and how to revoke it.