Security
Security
Filepad is built around the principle that agents — FilepadAI or connected external agents — should have exactly the access they need, no more, and that every higher-risk change should be reviewable before it lands. This page covers the security model at a product level. For the full technical reference, see the security overview in our docs.
Authentication
- Session-based authentication, with single sign-on via Microsoft Entra ID for enterprise environments
- Role-based access within workspaces: viewer, editor, admin, owner
- Workspace isolation enforced at the application layer — you only see workspaces you've been granted access to
Agent Access
External agents connect using OAuth-authenticated Agent Access keys, each scoped to a single workspace and to specific permissions:
- A key for one workspace cannot access another workspace
- Every request is checked against the scopes granted to that key before it touches any data
- Invalid or revoked tokens return a uniform 401 response, without revealing which condition failed
- Requests are subject to per-connection rate limiting, so a misbehaving or compromised agent can't overwhelm your workspace or run up costs unchecked
- Every key action — creation, rotation, scope change, revocation, and request — is logged with attribution and a timestamp
Workspace admins can create, rotate, patch scopes on, and revoke keys at any time from the Agents panel — rotation and scope changes don't require re-issuing the key.
Governed writes
The diff-and-approval model is a security control, not just a UX feature. Higher-risk actions from FilepadAI or a connected agent — editing an existing document, moving or deleting files — can pause and ask a human to approve them before they take effect. A small set of destructive actions always require approval, regardless of who requested them.
Data handling
- Filepad does not train its own models on your content
- Secrets — OAuth tokens, Agent Access keys, connector credentials — are encrypted at rest with AES-256-GCM
- Workspace isolation is enforced at the application layer on every request, not left to database configuration alone
See our Privacy Policy for how content is handled when it's sent to an AI provider for processing.
Reporting a security issue
Found a vulnerability? Email admin@filepad.ai with details. We take security reports seriously and will respond promptly.