Security & trust
Understanding an agent should not widen its attack surface.
Lanternfish treats credentials, telemetry, and workspace boundaries as product architecture—not a badge added after launch. This page states the controls in the product today and avoids claiming certifications we have not earned.
Security posture · Product controls
Opaque credentials
API keys and session tokens are stored as hashes. Raw credentials are shown only when they are created.
Workspace boundaries
Projects, members, usage, and billing limits resolve through the workspace boundary before data is returned.
Data minimization
Lanternfish is built to observe agent behavior. Teams should filter secrets and unnecessary personal data before telemetry leaves their environment.
01
Access and isolation
Dashboard access requires a verified local session and membership in the requested workspace. SDK ingestion uses project-scoped API keys. Authorization is resolved again at the API boundary instead of relying on the browser to hide data.
02
Retention and deletion
Managed Cloud reads and cleanup follow the active plan’s rolling retention window: 14 days on Free, 90 on Pro, 365 on Team, and a defined policy for Enterprise. Self-hosted operators control their own storage lifecycle.
03
Hosted AI boundaries
Hosted AI calls pass through workspace credits, idempotency, concurrency limits, and global circuit breakers. Failed internal or provider requests do not consume a customer credit. Self-hosted deployments choose their own provider and key.
04
Compliance status
Lanternfish does not currently claim SOC 2, ISO 27001, or another third-party certification on this page. Enterprise requirements are reviewed directly so promised controls and deployed controls stay aligned.
Responsible disclosure
Found something we should see?
Send reproduction steps and impact privately. Do not include active credentials or access data that is not yours.